Reply information generation method and apparatus, electronic device, and storage medium

Abstract

The present invention relates to the technical field of human-computer interaction. Disclosed are a reply information generation method and apparatus, an electronic device, and a storage medium. The specific implementation solution is: acquiring user interaction data; and generating reply information on the basis of the user interaction data, wherein the reply information at least comprises first reply content, and the information amount of the first reply content corresponds to the user interaction data. Embodiments of the present invention allow to flexibly increase or decrease the information amount of reply information, making it easy for a user to understand the reply information.

Description

Reply information generation method and device, electronic equipment and storage medium

[0001] Cross-reference to related applications

[0002] The present disclosure is based on and claims priority to Chinese Patent Application No. 202411099279.3, filed on August 9, 2024, Chinese Patent Application No. 202510040587.7, filed on January 9, 2025, Chinese Patent Application No. 202510040630.X, filed on January 9, 2025, and Chinese Patent Application No. 202510040441.2, filed on January 9, 2025, the contents of all of which are incorporated herein by reference in their entirety. TECHNICAL FIELD

[0003] The present application relates to the field of human-computer interaction, and in particular to a reply information generation method and device, an electronic equipment and a storage medium. BACKGROUND

[0004] With the development of vehicle technology, in the field of vehicle technology, users on the vehicle interact with the vehicle in a voice interaction manner. By recognizing the user's voice question, understanding the user's intention, determining the reply content according to the user's intention, and outputting the reply content by the vehicle.

[0005] At present, the reply content determined according to the user's intention is directly outputted, which cannot guide the user, and the reply content determined according to the user's intention is usually outputted according to the same information structure, resulting in poor human-computer interaction experience. The reply content determined according to the user's intention usually contains a large amount of information, and the vehicle directly outputs the reply content, which makes it difficult for the user to understand. SUMMARY

[0006] The present application provides a reply information generation method and device, an electronic equipment and a storage medium, which can flexibly increase or decrease the amount of reply information, and facilitate the user to understand the reply information.

[0007] In a first aspect, the present application provides a reply information generation method, comprising: obtaining user interaction data; generating reply information according to the user interaction data, wherein the reply information at least comprises first reply content, and an information amount of the first reply content corresponds to the user interaction data.

[0008] In a second aspect, the present application provides a reply information generation method, comprising: obtaining user interaction data; determining a guidance type corresponding to the user interaction data according to the user interaction data; generating reply information according to the user interaction data and the guidance type; wherein the reply information at least comprises first reply content, and whether the reply information comprises second reply content corresponds to the guidance type.

[0009] In a third aspect, the present application provides a reply information generation method, comprising: obtaining user interaction data; determining a reply information structure according to the user interaction data; and generating reply information according to the user interaction data and the reply information structure.

[0010] In a fourth aspect, the present application further provides a reply information generation device, comprising: an interaction data obtaining module, configured to obtain user interaction data; and a reply information generation module, configured to generate reply information according to the user interaction data, wherein the reply information at least comprises first reply content, and an information amount of the first reply content corresponds to the user interaction data.

[0011] In a fifth aspect, the present application further provides an electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the reply information generation method provided by any of the embodiments of the present application.

[0012] In a sixth aspect, the present application further provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to enable a processor to execute the reply information generation method of any of the embodiments of the present application.

[0013] The embodiments of the present application can flexibly adjust the information amount of the reply information by generating the reply information adapting to the information amount of the user interaction data, and specifically realize increasing or decreasing the information amount of the reply information, so as to feedback the reply information convenient for the user to understand.

[0014] It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiments description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0016] Fig. 1 is a flow chart of a reply information generation method according to an embodiment of the present application;

[0017] Fig. 2 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0018] Fig. 3 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0019] Fig. 4 is an application scenario diagram of a reply information generation method according to an embodiment of the present application;

[0020] Fig. 5 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0021] Fig. 6 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0022] Fig. 7 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0023] Fig. 8 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0024] Fig. 9 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0025] Fig. 10 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0026] Fig. 11 is a flow chart of another reply information generation method according to an embodiment of the present application;

[0027] Fig. 12 is a schematic diagram of a reply information generation device according to an embodiment of the present application;

[0028] Fig. 13 is a schematic diagram of another reply information generation device according to an embodiment of the present application;

[0029] Fig. 14 is a schematic diagram of another reply information generation device according to an embodiment of the present application;

[0030] Fig. 15 is a structural schematic diagram of an electronic device implementing the reply information generation method according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to better understand the technical scheme of the present application, the technical scheme in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first" and "second" and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to describe a specific order or a chronological order. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] In the technical scheme of the embodiments of the present application, the acquisition, storage and application of user information and the like are in line with the relevant legal regulations and do not violate public order and good customs.

[0034] FIG. 1 is a flowchart of a reply information generation method provided by an embodiment of the present application. The present embodiment can be applied to the case of replying to the user's question in the human-computer interaction process between the vehicle and the user. The method can be executed by a reply information generation device. The reply information generation device can be realized in the form of hardware and / or software, and is specifically configured in an electronic device. The electronic device can be a server or a terminal device. The terminal device can include a mobile phone, a computer, a vehicle-mounted terminal or a wearable device, etc.

[0035] Referring to the reply information generation method shown in FIG. 1, the method comprises the following steps.

[0036] S101, acquiring user interaction data.

[0037] In an optional embodiment, the user interaction data can be data related to the interaction in the process in which the human-computer interaction device interacts with the user. Illustratively, the human-computer interaction device can include a mobile phone, a computer, a vehicle terminal, or a wearable device, etc. In an example, the human-computer interaction device is a vehicle terminal. The vehicle interacts with the user, and the vehicle terminal obtains the question content input by the user, which can be used as the user interaction data. Meanwhile, the vehicle terminal can also obtain the user-related data, which can also be used as the user interaction data.

[0038] It should be noted that the reply information generation method of the embodiments of the present application can be executed by a server or a human-computer interaction device. When the server executes the reply information generation method of the embodiments of the present application, the user interaction data obtained by the server can be data obtained by processing the question data input by the user directly collected, for example, the human-computer interaction device processes the question data to obtain the question content and the user-related data, and determines the user interaction data according to the question content and the user-related data. When the human-computer interaction device executes the reply information generation method of the embodiments of the present application, the human-computer interaction device directly collects the question data input by the user to obtain data, directly uses the collected data as the user interaction data, or processes the collected data to use the processing result as the user interaction data. The human-computer interaction device can obtain the question data input by the user by at least one of the following ways: obtaining the text information input by the user, collecting the voice of the user, taking the image of the user, or recording the video of the user.

[0039] In an optional embodiment, the human-computer interaction device that interacts with the user is a vehicle. In the process in which the vehicle interacts with the user, the vehicle terminal generates interaction data related to the human-computer interaction operation, and determines the user interaction data according to the generated interaction data.

[0040] S102, generating reply information according to the user interaction data, the reply information at least including first reply content, the information amount of the first reply content corresponding to the user interaction data.

[0041] In an optional embodiment, the reply information can be content for replying to the user interaction data. The information amount can be a numerical value obtained by quantifying the reply information. Illustratively, the data amount included in the reply information can be used to represent, for example, the data amount can be the number of bytes, the number of characters, or the number of words (characters), etc. Alternatively, the data amount of the effective data in the reply information can also be used to represent the information amount of the reply information, for example, the number of entities included in the reply information is used as the information amount in the reply information, and for example, the data amount of the keywords included in the reply information is used as the information amount in the reply information.

[0042] In an optional embodiment, the user interaction data is used to determine the substantial content of the reply information, for example, the user interaction data is used to determine the answer of the user interaction data. The user interaction data is also used to determine the information amount of the first reply content. The first reply content is the substantial content of the reply information, which can be understood as the answer of the user interaction data, and the information amount of the first reply content is determined by the user interaction data.

[0043] In an optional embodiment, the generated reply information is generally text data, which can be sent to the human-computer interaction device, and the human-computer interaction device displays the reply information in the form of text to the user; or the generated reply information is an image displayed to the user; the generated reply information is corresponding voice played to the user; or the generated reply information is corresponding video played to the user, etc. Alternatively, according to the reply information, at least one of the corresponding image, the corresponding voice or the corresponding video is generated, and at least one of the image, the voice and the video is sent to the human-computer interaction device.

[0044] In an example, the user's question content is: why did the dinosaurs disappear? For example, the reply information only includes the first reply content with a large amount of information, and the first reply content is: wow, this is a very good question! The dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky and hitting the earth with a loud noise, which is a big meteorite. This big stone caused a lot of dust and smoke to block the sky, and the sun's light could not reach the ground, and the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0045] For another example, the reply content includes the first reply content with a small amount of information. The first reply content is: this is a very good question, the disappearance of dinosaurs is due to the impact of a small asteroid on the earth, which causes a dramatic change in the environment, leading to the inability to survive, but the process is also complex and interesting.

[0046] For example, according to the user interaction data, the reply information can be generated by inputting the user interaction data into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data can be filled into a prompt template to obtain input data, and the input data is input into a large language model to obtain the reply information output by the large language model.

[0047] For another example, according to the user interaction data, the reply information can be generated by generating reply instruction information according to the user interaction data, and generating the reply information according to the reply instruction information and the user interaction data.

[0048] In an optional embodiment, the reply indication information is used to determine the information amount of the reply information, and the reply indication information and the user interaction data are used to cooperate to generate the reply information meeting the information amount corresponding to the reply indication information.

[0049] For example, the reply indication information can be a type or a label. Specifically, the reply indication information can include a type of small information amount and no guiding follow-up question, a type of small information amount and guiding follow-up question, a type of large information amount and no guiding follow-up question, and a type of large information amount and guiding follow-up question.

[0050] For another example, the reply indication information can be input data of a model. Specifically, the reply indication information can be: please provide an answer that a child can understand for a question of a child: XX, and add guiding content meeting the interest of the child. For another example, the reply indication information can be: please provide a complete answer for a question of an adult: XX. In addition, the reply indication information can have other description manners, which are not specifically limited.

[0051] Specifically, according to the user interaction data, the reply indication information is generated, and according to the reply indication information, the reply information corresponding to the user interaction data is generated. The method can include: inputting the user interaction data into a first large language model to obtain reply indication information output by the first large language model, wherein the reply indication information can be a type or a label; querying a prompt template corresponding to the reply indication information according to the reply indication information, filling the user interaction data into the corresponding prompt template to obtain input data, and inputting the input data into a second large language model to obtain reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0052] For another example, the user interaction data is input into a first large language model to obtain reply indication information output by the first large language model, and the reply indication information is a prompt template. The user interaction data is filled into the reply indication information to obtain input data, and the input data is input into a second large language model to obtain reply information output by the second large language model. The first large language model and the second large language model can be the same or different.

[0053] For another example, the user interaction data is input into a first large language model to obtain reply indication information output by the first large language model, wherein the reply indication information can be a type or a label; the reply indication information and the user interaction data are input into a second large language model to obtain reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0054] The reply indication information is obtained by processing the user interaction data, and the reply information is generated according to the reply indication information and the user interaction data. The effective data affecting the information amount in the user interaction data can be further extracted, and the reply indication information is generated, so that the reply information is generated based on the reply indication information and the user interaction data, the accuracy of the reply information can be improved, the flexibility of the reply information can be increased, and the load pressure caused by end-to-end of the reply information generated only according to the user interaction data can be reduced.

[0055] The first reply content is generated according to the information amount of the user interaction data, and the reply information is determined according to the first reply content, so that the information amount of the reply information can be flexibly adjusted, and the information amount of the first reply content is specifically realized, thereby facilitating the user to understand the reply information.

[0056] In some optional embodiments, the reply information is generated according to the user interaction data, including: at least one of the target information amount, the guide type, and the reply information structure corresponding to the user interaction data is determined according to the user interaction data; and the reply information is generated according to at least one of the target information amount, the guide type, and the reply information structure and the user interaction data.

[0057] FIG. 2 is a flowchart of another reply information generation method provided by an embodiment of the present application.

[0058] In an optional embodiment, the step of generating the reply information according to the user interaction data is refined as: the target information amount corresponding to the user interaction data is determined according to the user interaction data; and the reply information is generated according to the user interaction data and the target information amount, and the information amount of the first reply content corresponds to the target information amount.

[0059] In some optional embodiments, the step of determining at least one of the target information amount, the guide type, and the reply information structure corresponding to the user interaction data according to the user interaction data includes:

[0060] At least one of the target information amount, the guide type, and the reply information structure corresponding to the user interaction data is determined according to at least one of the user attribute information, the difficulty level corresponding to the user question content, the known information corresponding to the user historical dialogue, the preferred content corresponding to the user behavior feature, and the operation complexity corresponding to the vehicle data.

[0061] It should be noted that the parts not described in detail in the embodiments of the present application can be referred to the descriptions of the foregoing embodiments.

[0062] Referring to the reply information generation method shown in FIG. 2, the method includes:

[0063] S201, obtaining user interaction data.

[0064] S202, determine a target information amount corresponding to the user interaction data according to the user interaction data.

[0065] In an optional embodiment, the target information amount is used to indicate an information amount of the generated first reply content. The target information amount can be a specific numerical value, for example, the number of words. Or the target information amount can be a level or type. For example, the target information amount is more or less information amount.

[0066] In an optional embodiment, the manner of determining the target information amount can be, for example, a mapping relationship between the user interaction data and the target information amount can be preset, and the target information amount corresponding to the user interaction data is queried according to the mapping relationship. For example, the user interaction data can be input into a pre-trained deep learning model to obtain the target information amount. Specifically, the user interaction data is feature extracted to obtain a feature vector, the feature vector is decoded and classified to obtain the target information amount. For example, the user interaction data is quantized to obtain the target information amount.

[0067] S203, generate reply information according to the user interaction data and the target information amount, and the information amount of the first reply content corresponds to the target information amount.

[0068] In an optional embodiment, the target information amount corresponding to the information amount of the first reply content can mean that the target information amount is the same as the information amount of the first reply content, or the type of the target information amount is the same as the type of the information amount of the first reply content. For example, the type of the target information amount and the type of the information amount of the first reply content are both the type of more information amount. For example, the type of the target information amount and the type of the information amount of the first reply content are both the type of less information amount.

[0069] In an optional embodiment, generating reply information according to the user interaction data and the target information amount can be: inputting the user interaction data and the target information amount into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data and the target information amount can be filled into a prompt template corresponding to the target information amount to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0070] For example, according to the user interaction data and the target information quantity, generating the reply information can be: according to the user interaction data and the target information quantity, generating reply indication information, and according to the reply indication information and the user interaction data, generating the reply information. Specifically, input the user interaction data into the first large language model to obtain the target information quantity output by the first large language model, and query or generate the reply indication information according to the target information quantity. According to the user interaction data and the reply indication information, determine the input data. Input the input data into the second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0071] The embodiment of the application can obtain the target information quantity by processing the user interaction data, and generate the reply information including the first reply content corresponding to the target information quantity according to the target information quantity and the user interaction data. The target information quantity can be directly extracted from the user interaction data, and the reply information can be generated based on the target information quantity and the user interaction data. The information quantity of the first reply content can be accurately controlled.

[0072] Optionally, the user interaction data includes user attribute information. The user attribute information can be information that is stable and unchangeable for a user. The user attribute information can include: user identification (identity), gender, age, height, and cognitive level, etc. For example, the user's voiceprint can be recognized through the user's voice, and the user's voiceprint can uniquely identify the user. For another example, the user's face features can be recognized through the user's image, and the user's face features can uniquely identify the user. Generally, the user interacts with the vehicle through voice, and the user identification can be represented by the voiceprint features recognized by the user's voice. In addition, the user's voice, image, and video can be used to identify the user's gender, age, and cognitive level. The user can also directly input the user identification, gender, age, height, and cognitive level, etc. in text.

[0073] In an optional embodiment, the user attribute information can be obtained by at least one of the following ways: receiving text information input directly by the user, collecting the user's voice, collecting the user's image, and collecting the user's video. Correspondingly, the media type of the user information obtained by collecting the user's voice is audio type; the media type of the user information obtained by receiving the text data input by the user is text type; the media type of the user information obtained by collecting the user's image is image type; and the media type of the user information obtained by collecting the user's video is video type.

[0074] In an optional embodiment, the user's understanding ability and / or preferred content can be determined according to the user attribute information, so as to determine the target information amount according to the user's understanding ability and / or preferred content.

[0075] For example, the user attribute information includes that the user's occupation is an athlete, and the user's question content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in the art work, and the target information amount is determined to be small.

[0076] For another example, the user attribute information includes that the user's occupation is an e-sports player, and the user's question content is the introduction of XX game. It is determined that the user has a high interest in the game, and the target information amount is determined to be large.

[0077] For another example, the user attribute information includes that the user's age is a child's age. It is determined that the user's understanding ability is weak, and the target information amount is determined to be small.

[0078] Specifically, the user attribute information can include the user's age. For example, when the user's age is less than or equal to a preset age threshold, the target information amount is determined to be small, and the first reply content with small information amount is generated; when the user's age is greater than the preset age threshold, the target information amount is determined to be large, and the first reply content with large information amount is generated.

[0079] Specifically, the user attribute information can include the user's identity, for example, when the user's identity is outside a preset identity range, the target information amount is determined to be small, and the first reply content with small information amount is generated; when the user's identity belongs to the preset identity range, the target information amount is determined to be large, and the first reply content with large information amount is generated.

[0080] Specifically, the user attribute information can include the user's gender, for example, when the user's gender is a first gender, the target information amount is determined to be small, and the first reply content with small information amount is generated; when the user's gender is a second gender, the target information amount is determined to be large, and the first reply content with large information amount is generated.

[0081] Specifically, the user attribute information can include the user's height, for example, when the user's height is less than or equal to a preset height threshold, the target information amount is determined to be small, and the first reply content with small information amount is generated; when the user's height is greater than the preset height threshold, the target information amount is determined to be large, and the first reply content with large information amount is generated.

[0082] Specifically, the user attribute information can include the user's cognitive level, for example, when the user's cognitive level is lower than a preset cognitive level, the target information amount is determined to be small, and the first reply content with small information amount is generated; when the user's cognitive level is higher than the preset cognitive level, the target information amount is determined to be large, and the first reply content with large information amount is generated.

[0083] Optionally, the user attribute information can further include occupation and other contents, which are not specifically limited.

[0084] In addition, when the user attribute information includes multiple items of user identification, gender, age, height, and cognitive level, it can be selected whether each item meets the corresponding condition to generate the first reply content with a corresponding amount of information. For example, the weighted score of each item can be calculated, and the first reply content with a corresponding amount of information can be generated according to the weighted score.

[0085] In one example, the user attribute information includes user age. The preset age threshold is 12 years old. When the user age is less than 12 years old, the target information amount is determined to be small, and the first reply content with a small amount of information is generated as: This is a very good question. The disappearance of dinosaurs is due to the impact of a small asteroid on the earth, which causes a dramatic change in the environment, leading to the inability to survive, but the process is also very complex and interesting. When the user age is greater than or equal to 12 years old, the target information amount is determined to be large, and the first reply content with a large amount of information is generated as: Wow, this is a very good question! Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge rock flying from the sky and hitting the earth with a loud bang, which is a large meteorite. This large stone caused a lot of dust and smoke to block the sky, and the sun's light could not reach the ground, making the earth very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0086] Optionally, the determining of the target information amount corresponding to the user interaction data according to the user interaction data includes: when the user attribute information is first user attribute information, determining a first information amount corresponding to the first user attribute information according to the first user attribute information; when the user attribute information is second user attribute information, determining a second information amount corresponding to the second user attribute information according to the second user attribute information; the first user attribute information is different from the second user attribute information, and the first information amount is different from the second information amount.

[0087] In one optional embodiment, the target information amount is different for different users with different user attribute information. Specifically, the user attribute information of a first user is first user attribute information, and the target information amount is determined to be a first information amount according to the first user attribute information. The user attribute information of a second user is second user attribute information, and the target information amount is determined to be a second information amount according to the second user attribute information. When the first user attribute information and the second user attribute information are different, the first information amount and the second information amount are different. For example, a mapping relationship between the user attribute information and the target information amount can be preset, and the corresponding target information amount can be determined according to the difficulty level. For example, the user attribute information can be input into a deep learning model to obtain the target information amount.

[0088] By specifically defining the user interaction data as the user attribute information, the understanding ability and the preferred content of the user can be considered, the information amount adapted to the understanding ability and the preferred content of the user can be determined, the first reply content convenient for the user to understand can be generated, and the first reply content in which the user is interested can be generated.

[0089] In an optional embodiment, the first user attribute information includes a first age, the second user attribute information includes a second age, the first age is different from the second age, and a first information amount corresponding to the first age is different from a second information amount corresponding to the second age. In an example, the understanding ability of a user with a small age is weak, and the understanding ability of a user with a large age is strong. The user with the small age can be fed back with content with a small information amount, so as to reduce the understanding difficulty of the reply content. The first age is smaller than the second age, and accordingly, the first information amount is smaller than the second information amount. In an example, the user with the small age can be fed back with content with a large information amount and interesting, so as to make the user with the small age understand the reply content. By specifically defining the user attribute information as the user age, the understanding ability of the user can be determined based on the age of the user, so as to determine the information amount adapted to the understanding ability of the user with the age, so as to generate the reply information corresponding to the target information amount, and the reply information convenient for the user to understand based on the age of the user can be provided.

[0090] Optionally, the user interaction data includes user question content. The user question content can refer to content input by the user to the human-computer interaction device for asking the human-computer interaction device in the process of the user interacting with the human-computer interaction device. Optionally, the determining, according to the user interaction data, of the target information amount corresponding to the user interaction data includes: when a difficulty level corresponding to the user question content is a first difficulty level, determining, according to the user question content, a third information amount corresponding to the user interaction data; when the difficulty level corresponding to the user question content is a second difficulty level, determining, according to the user question content, a fourth information amount corresponding to the user interaction data; and when the first difficulty level is different from the second difficulty level, the third information amount is different from the fourth information amount.

[0091] In an optional embodiment, the difficulty level corresponding to the user question content is used to determine the complexity of an answer, so as to determine the target information amount. The user question content can be classified according to the difficulty level, for example, a difficult degree or a simple degree. For example, a mapping relationship between the difficulty level and the target information amount can be preset, and the target information amount corresponding to the difficulty level can be determined according to the difficulty level. For another example, the user question content can be input into a deep learning model to obtain the target information amount.

[0092] In an optional embodiment, the target information quantity is different for different difficulty levels of user query content. The same user can input user query content of different difficulty levels. Difficult user query content can increase the target information quantity, implement feedback using first reply content with more information quantity, and facilitate user understanding of the first reply content; simple user query content can reduce the target information quantity, implement feedback using first reply content with less information quantity, and facilitate user quick understanding of the first reply content.

[0093] By specifically defining the user interaction data as user query content, the information quantity can be determined according to the difficulty level of the user query content, the information quantity of the query content can be adapted, and the first reply content that is convenient for the user to understand can be generated.

[0094] Optionally, the user interaction data includes user historical dialogues. The user historical dialogues can refer to historical dialogue information of the user related to the user interaction data. In fact, the embodiment of the application can be applied in a multi-turn dialogue scenario, and historical turns of dialogues of the user can be stored as user historical dialogues. In addition, all dialogues of the user sent by the man-machine interactive device before the current time can also be recorded as historical dialogues.

[0095] In an optional embodiment, the electronic device implementing the reply information generation method can record data of the user, and can locally query user-related data. For example, multiple sessions of the same user can be recorded, and one session can include at least one turn of dialogue. The user identifier can be stored in correspondence with the session of the user, and a corresponding correspondence relationship can be established. The user historical session can be queried according to the user identifier in the user attribute information, so as to obtain the user historical dialogues.

[0096] Optionally, the determining, according to the user interaction data, of a target information quantity corresponding to the user interaction data includes: when known information corresponding to the user historical dialogues is first known information, determining, according to the user historical dialogues, a fifth information quantity corresponding to the user interaction data; when the known information corresponding to the user historical dialogues is second known information, determining, according to the user historical dialogues, a sixth information quantity corresponding to the user interaction data; and the fifth information quantity and the sixth information quantity are different when the first known information and the second known information are different.

[0097] In an optional embodiment, the user historical conversation is used to determine the known information, so as to determine the target information amount according to the known information. The key information extracted from the user historical conversation can be taken as the known information. The target information amount is different for different user historical conversations. In fact, the first reply content does not need to repeat the content that has been replied in the user historical conversation. Therefore, the known information corresponding to the first reply content can be determined according to the user historical conversation. The target information amount is determined according to the known information. For example, a mapping relationship between the known information and the target information amount can be preset, for example, a mapping relationship between the information amount of the known information and the target information amount, for example, the mapping relationship is inversely proportional. The corresponding target information amount is determined according to the known information. For another example, the known information can be input into a deep learning model to obtain the target information amount.

[0098] The target information amount determined is different for different known information. The same user question content can correspond to the user historical conversation of different known information, and different users correspond to the user historical conversation of different known information. In an optional embodiment, the preference and understanding ability of the user can be determined according to the user historical conversation, and the information amount is determined according to the preference and understanding ability of the user.

[0099] By specifically limiting the user interaction data to the user historical conversation, the known information can be considered, so as to reduce the repeated known content, provide effective new reply content, and reduce redundant interaction.

[0100] Optionally, the user interaction data includes user behavior characteristics. The user behavior characteristics are used to determine the preference content of the user and the like. The user behavior characteristics can be determined according to the user historical conversation and the user attribute information and the like. Exemplarily, the user behavior characteristics indicate that the user historically prefers the content. The user behavior characteristics are queried according to the user identifier in the user attribute information. The user behavior characteristics can be the type or label of the user.

[0101] Optionally, the determining the target information amount corresponding to the user interaction data according to the user interaction data includes: when the preference content corresponding to the user behavior characteristics is first preference content, determining a seventh information amount corresponding to the user interaction data according to the user behavior characteristics; when the preference content corresponding to the user behavior characteristics is second preference content, determining an eighth information amount corresponding to the user interaction data according to the user behavior characteristics; and the seventh information amount and the eighth information amount are different when the first preference content and the second preference content are different.

[0102] In an optional embodiment, the user behavior characteristics are used to determine the preference content. The target information amount is determined according to the preference content.

[0103] The target information quantity is different for different preference contents. In fact, the user would like more interesting contents. Therefore, the target information quantity can be determined according to the preference content.

[0104] For example, a mapping relationship between the similarity between the preference content and the user query content and the target information quantity can be preset, the similarity between the preference content and the user query content is detected, and the target information quantity is queried according to the similarity and the corresponding relationship. For another example, the preference content and the user query content can be input into a deep learning model to obtain the target information quantity.

[0105] The target information quantity determined is different for different preference contents. The same user query content can correspond to different preference contents, and different users correspond to different preference contents.

[0106] In an optional embodiment, the more similar the preference content and the user query content are, the more the target information quantity is, and the more the preference content and the user query content deviate, the less the target information quantity is. For example, the user behavior characteristics include strong curiosity, or the user historical dialogue includes multiple multi-round dialogues, and it is determined that the information quantity is large; for another example, the user query content is a game-related question, the user behavior characteristics include a preference for astronomy, or the user historical dialogue includes multiple one-round simple dialogues, and it is determined that the information quantity is small, and the second reply content is not included.

[0107] In an optional embodiment, the user historical dialogue and / or the user behavior characteristics can be queried according to the user identifier in the user attribute information to obtain a query result; and the user interaction data is generated according to the query result, the user attribute information and the user query content.

[0108] By limiting the user interaction data to the user behavior characteristics, determining the preference content according to the user behavior characteristics, and determining the target information quantity, more reply contents can be provided for the user's interesting questions.

[0109] Optionally, the user interaction data further includes vehicle data. In an optional embodiment, the human-computer interaction device can be a vehicle, and the vehicle data can be vehicle-related data. The vehicle data can include vehicle state, vehicle-machine state and extravehicular environment, etc. The vehicle state includes but is not limited to at least one of vehicle speed, tire pressure and vehicle temperature, etc.; the vehicle-machine state can include but is not limited to at least one of foreground application usage state and background application state, etc.; the extravehicular environment state includes but is not limited to at least one of weather, GPS positioning data and road surface condition, etc.

[0110] In an optional embodiment, the vehicle data is used as a basis for a reply to the user query content related to the vehicle. For example, the user query content is: How long will the vehicle arrive? The speed of the vehicle, the driving route of the vehicle, the current position of the vehicle, the real-time road conditions on the driving route, and the current time can be obtained as vehicle data, and the arrival time and the arrival time of the vehicle are calculated according to the vehicle data. The vehicle data is used to determine the substantive content of the first reply content in combination with the user query content, and the substantive content can be an answer.

[0111] In an optional embodiment, the vehicle information can include vehicle attribute information obtained by the vehicle locally and vehicle status collected by the vehicle in real time, such as the model of the vehicle, the brand of the vehicle, the size of the vehicle, or the structure of the vehicle, etc. Among them, the vehicle status can be further divided into vehicle itself status, car machine status, and off-cabin environment status, etc. More specifically, the vehicle status can include but is not limited to at least one of the vehicle speed, the tire pressure, and the temperature in the vehicle; the car machine status can include but is not limited to at least one of the foreground application usage status and the background application status; and the off-cabin environment status includes but is not limited to at least one of the weather, the GPS (Global Positioning System) positioning data, and the road surface condition.

[0112] It should be noted that the acquisition, storage, and application of user information involved in the embodiments of the present application all comply with relevant laws and regulations, do not violate public order and good customs, and are authorized by the user. The vehicle will only acquire the user information, and the user information transmitted by the vehicle is processed data. The vehicle will not transmit the source data of the user information to the electronic device implementing the reply information generation method.

[0113] In an example, the first reply content with less information can be: The vehicle will arrive in XX minutes, and is expected to arrive at 10:32. For another example, the first reply content with more information can be: The current vehicle has traveled to XX place, and there are XX meters to the end. The road ahead is smooth, the vehicle will arrive in XX minutes, and is expected to arrive at 10:32.

[0114] By limiting the user interaction data to vehicle data and limiting the interaction scene to the human-vehicle interaction scene, the input data can be enriched based on the user query content, making the input data more diverse. Based on more complete input data, the reply information can be generated, the reply information in the human-vehicle interaction in the vehicle scene can be increased or decreased, and the user can be helped to understand the real-time status of the vehicle more quickly.

[0115] Optionally, the determining the target information amount corresponding to the user interaction data according to the user interaction data comprises: when the operation complexity degree corresponding to the vehicle data is a first operation complexity degree, determining a ninth information amount corresponding to the user interaction data according to the vehicle data; when the operation complexity degree corresponding to the vehicle data is a second operation complexity degree, determining a tenth information amount corresponding to the user interaction data according to the vehicle data; when the first operation complexity degree is different from the second operation complexity degree, the ninth information amount is different from the tenth information amount.

[0116] The operation complexity degree is used to describe the complexity of processing the first reply content. The user query content and the vehicle data can be used to determine the vehicle data related to the user query content, classify the related vehicle data, and determine the operation complexity degree.

[0117] The target information amount determined is different for different operation complexity degrees. Generally, the more vehicle data needed by the user query content, the more complex the process of processing the first reply content, and the more first reply content provided, i.e., the target information amount is large; the less vehicle data needed by the user query content, the simpler the process of processing the first reply content, and the less first reply content provided, i.e., the target information amount is small. Different user query contents can correspond to different operation complexity degrees. Generally, the target information amount of high operation complexity degree is large, and the target information amount of low operation complexity degree is small. For example, a mapping relationship between the operation complexity degree and the target information amount can be preset, and the corresponding target information amount can be determined according to the operation complexity degree. For another example, the operation complexity degree can be input into a deep learning model to obtain the target information amount.

[0118] By specifically limiting the user interaction data to the vehicle data, the complexity of processing the vehicle data in the human-vehicle interaction scenario can be considered to determine the information amount, the information amount can be adapted to the complexity of processing the vehicle data, and the first reply content convenient for the user to understand can be generated.

[0119] In an optional embodiment, the user interaction data can include user query content and user attribute information.

[0120] In an example, the user query content of the user is: why did the dinosaurs disappear? The user attribute information includes the user age, and the user age is 10 years old. Correspondingly, the first reply content with a small amount of information is: this is a very good question, the disappearance of dinosaurs is due to the impact of a small planet on the earth, which causes a dramatic change in the environment, leading to an inability to survive, but the process is also very complex and interesting.

[0121] In one example, the user's question content is: why did dinosaurs disappear? The user attribute information includes the user's age, and the user's age is 20 years old. Accordingly, the first reply content with a large amount of information is: Dinosaurs disappeared because a long time ago, a huge rock came from the sky and crashed onto the earth, causing a lot of dust and smoke to block the sky, the sun's light could not reach the ground, the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0122] In an optional embodiment, the user interaction data can include user question content, user attribute information, and user historical dialogue. In one example, the user's question content is: why is the speed of A greater than the speed of B? The user's historical dialogue is the content of the user's first question: who has a faster speed between A and B, A travels S1 at T1, and B travels S2 at T2? The first reply content is: A's speed is greater than B's speed. The user attribute information includes the user's age, and the user's age is 12 years old. Accordingly, the first reply content with a small amount of information is: S1 / T1 = V1, therefore, V1 > V2. In one example, the user's question content is: why is the speed of A greater than the speed of B? The user's historical dialogue is the content of the user's first question: who has a faster speed between A and B, A travels S1 at T1, and B travels S2 at T2? The first reply content is: A's speed is greater than B's speed. The user attribute information includes the user's age, and the user's age is 12 years old. Accordingly, the first reply content with a large amount of information is: S1 / T1 = V1, S2 / T2 = V2, V1 > V2.

[0123] In an optional embodiment, the user interaction data can include user question content, user attribute information, and user behavior characteristics. In one example, the user's question content is: why did dinosaurs disappear? The user attribute information includes the user's age, and the user's age is 10 years old. The user behavior characteristics are that the user likes cute and cute things. Accordingly, the first reply content with a small amount of information is: This is a very good question, the disappearance of dinosaurs is due to the impact of a small asteroid on the earth, causing a dramatic change in the environment, leading to an inability to survive, but the process is also complex and interesting. In one example, the user's question content is: why did dinosaurs disappear? The user attribute information includes the user's age, and the user's age is 10 years old. The user behavior characteristics are that the user likes dinosaurs. Accordingly, the first reply content with a large amount of information is: Dinosaurs disappeared because a long time ago, a huge rock came from the sky and crashed onto the earth, causing a lot of dust and smoke to block the sky, the sun's light could not reach the ground, the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0124] In an optional embodiment, the user interaction data can include user asking content, user attribute information, user historical dialogue, and user behavior characteristics. In an example, the user asking content is: why did dinosaurs disappear? The user attribute information includes user age, and the user age is 10 years old. The user historical dialogue includes related questions about dolls, and the user behavior characteristics are liking cute and cute things. Accordingly, the first reply content with a small amount of information is: this is a very good question, the disappearance of dinosaurs is because a small asteroid hit the earth, causing a dramatic change in the environment, leading to the inability to survive, but the process is also very complex and interesting. In an example, the user asking content is: why did dinosaurs disappear? The user attribute information includes user age, and the user age is 10 years old. The user historical dialogue includes related questions about Tyrannosaurus rex, and the user behavior characteristics are liking dinosaurs. Accordingly, the first reply content with a large amount of information is: dinosaurs disappeared because a long time ago, a huge stone came from the sky and hit the earth with a loud crash, causing a lot of dust and smoke to block the sky, the sun's light could not reach the ground, the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0125] In an optional embodiment, the user interaction data can include user asking content, user attribute information, and vehicle data. In an example, the user asking content is: how far is the service area? The user attribute information includes user age, and the user age is 10 years old. According to the location of the vehicle in the vehicle data, the nearest service area is determined, the distance between the vehicle and the service area is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the service area. Accordingly, the first reply content with a small amount of information is: it will take 20 minutes to arrive. In an example, the user asking content is: how far is the service area? The user attribute information includes user age, and the user age is 20 years old. According to the location of the vehicle in the vehicle data, the nearest service area is determined, the distance between the vehicle and the service area is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the service area. Accordingly, the first reply content with a large amount of information is: the distance between the nearest service area and the current location is 20 km, and it is expected to take 30 minutes to arrive at the current speed.

[0126] In an optional embodiment, the user interaction data can include user asking content, user attribute information, user historical dialogue, and vehicle data. In an example, the user asking content is: How far is the restaurant? The user attribute information includes user age, and the user age is 10 years old. The user historical dialogue is the content of the first asking of the user: Is there a noodle shop nearby? The first reply content is: There is a noodle shop. According to the location of the vehicle in the vehicle data, the nearest noodle shop is determined, the distance between the vehicle and the noodle shop is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the noodle shop. Accordingly, the first reply content with less information is: It will take 20 minutes to arrive. In an example, the user asking content is: How far is the restaurant? The user attribute information includes user age, and the user age is 20 years old. The user historical dialogue is the content of the first asking of the user: Is there a noodle shop nearby? The first reply content is: There is a noodle shop. According to the location of the vehicle in the vehicle data, the nearest noodle shop is determined, the distance between the vehicle and the noodle shop is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the noodle shop. Accordingly, the first reply content with more information is: The distance between the nearest noodle shop and the current position is 20 km, and it is expected to take 30 minutes to arrive at the constant speed according to the current vehicle speed.

[0127] In an optional embodiment, the user interaction data can include user asking content, user attribute information, user behavior characteristics, and vehicle data. In an example, the user asking content is: What is fun nearby? The user attribute information includes user age, and the user age is 10 years old. The user behavior characteristics are: like swimming. According to the location of the vehicle in the vehicle data, the nearest swimming pool is determined, the distance between the vehicle and the swimming pool is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the swimming pool. Accordingly, the first reply content with less information is: There is a swimming pool 1 km away. In an example, the user asking content is: What is fun nearby? The user attribute information includes user age, and the user age is 20 years old. The user behavior characteristics are: like swimming. The user behavior characteristics are: like swimming. According to the location of the vehicle in the vehicle data, the nearest swimming pool is determined, the distance between the vehicle and the swimming pool is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the swimming pool. Accordingly, the first reply content with more information is: There is a swimming pool 1 km away, and it is expected to take 10 minutes to arrive.

[0128] In an optional embodiment, the user interaction data can include user asking content, user attribute information, user historical dialogue, user behavior characteristics, and vehicle data. In an example, the user asking content of the user is: how long can we reach a fun place. The user attribute information includes the user age, and the user age is 10 years old. The user historical dialogue is the content of the first asking of the user: what fun places are nearby? The first reply content is: there are swimming pool, basketball court and badminton court. The user behavior characteristics are: like swimming. According to the location of the vehicle in the vehicle data, the nearest swimming pool is determined, the distance between the vehicle and the swimming pool is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the swimming pool. Accordingly, the first reply content with less information is generated as: there is a swimming pool 1 km away. In an example, the user asking content of the user is: what fun places are nearby? The user attribute information includes the user age, and the user age is 20 years old. The user historical dialogue is the content of the first asking of the user: what fun places are nearby? The first reply content is: there are swimming pool, basketball court and badminton court. The user behavior characteristics are: like playing basketball. According to the location of the vehicle in the vehicle data, the nearest basketball court is determined, the distance between the vehicle and the swimming pool is obtained, and the driving time is calculated according to the current speed of the vehicle in the vehicle data and the distance between the vehicle and the swimming pool. Accordingly, the first reply content with more information is generated as: there is a basketball court 2 km away, and it is expected to take 20 minutes to reach. And there is a swimming pool 1 km away, and it is expected to take 10 minutes to reach. There is no badminton court nearby.

[0129] In an optional embodiment, the user interaction data can include at least one of the following: user attribute information, user historical dialogue, user behavior characteristics, vehicle data, and user asking content. When the user interaction data includes at least two contents, according to the user interaction data, the target information quantity corresponding to the user interaction data can be: for each content, determining the information quantity corresponding to each content; weighting the information quantity determined for each content to obtain a weighting result; and taking the weighting result as the target information quantity corresponding to the user interaction data. In an example, the user interaction data of the first user includes at least one of the following: first user attribute information, first user asking content with first difficulty level, first user historical dialogue with first known information, first user behavior characteristics with first preferred content, and first vehicle data with first operation complexity. At least one of the first information quantity, the third information quantity, the fifth information quantity, the seventh information quantity, and the ninth information quantity can be weighted and calculated to obtain the target information quantity corresponding to the user interaction data.

[0130] In addition, for the same user asking content, the first reply content corresponding to the user interaction data of different combinations of contents is different.

[0131] The content of the user interaction data is flexibly combined, the flexibility of the reply content is increased, and various interaction requirements and various interaction scenarios are met.

[0132] FIG. 3 is a flow of another reply information generation method provided by an embodiment of the present application.

[0133] In an optional embodiment, the reply information further includes second reply content.

[0134] It should be noted that the parts not described in detail in the embodiments of the present application can refer to the descriptions of the foregoing embodiments.

[0135] Referring to the reply information generation method shown in FIG. 3, the method includes the following steps.

[0136] S301, obtaining user interaction data.

[0137] S302, generating reply information according to the user interaction data, the reply information at least including first reply content, the information amount of the first reply content corresponding to the user interaction data, the reply information further including second reply content, the second reply content being guiding follow-up question content corresponding to the user interaction data.

[0138] In an optional embodiment, the second reply content can be optional. The reply information can include the second reply content, or can not include the second reply content. Whether the reply information includes the second reply content is determined by the user interaction data. As in the foregoing embodiment, whether the reply information includes the second reply content can also be determined by the reply indication information determined according to the user interaction data.

[0139] In fact, when the reply information only includes the first reply content with a small amount of information, when the curiosity of the user is relatively strong, or when the user is relatively interested, the user can be prompted to continue to ask follow-up questions about more and richer content outside the reply information. The specific prompting manner can be to supplement guiding follow-up question content in the reply information to prompt the user to continue to ask follow-up questions about more and richer content outside the reply information of the current question content. The guiding follow-up question content is used to guide the user to continue to ask follow-up questions about the question content, and the guiding follow-up question content is content for guiding the user to continue to ask follow-up questions about the question content on the basis of the first reply content. The guiding follow-up question content can include suggestion information, counter-question information, and recommendation information, etc.

[0140] In an example, as in the foregoing embodiment, the reply information not including the second reply content is: the main reason for the disappearance of dinosaurs is asteroid impact on the earth, leading to environmental changes (first reply content).

[0141] For another example, the reply information including the second reply content is: the main reason for the disappearance of dinosaurs is asteroid impact on the earth, leading to environmental changes (first reply content). Do you want to know the story of this big meteorite impact on the earth (second reply content)?

[0142] By determining whether the second reply content is included in the reply information through the user interaction data, it can be convenient for the user to select whether to obtain the reply information with complete information amount, and by guiding the second reply content of follow-up questions, the first reply content with flexible increase and decrease of information amount can be matched, the user can be prompted and guided to continue to ask questions to obtain complete reply content, so as to make the user ask questions gradually, which conforms to the real dialogue scene, thereby improving the authenticity of the reply information and improving the user experience.

[0143] In an optional embodiment, the generating the reply information according to the user interaction data comprises: determining a reply information structure according to the user interaction data; and generating the reply information according to the user interaction data and the reply information structure.

[0144] In an optional embodiment, the reply information structure is used to represent the inclusion of different types of reply content. Different types of reply content can include: first reply content and second reply content.

[0145] In an optional embodiment, the reply information structure includes the first reply content. In an optional embodiment, the reply information structure can also include the first reply content and the second reply content. Whether the second reply content is included in the reply information structure is determined by the user interaction data.

[0146] In an optional embodiment, the user interaction data is used to determine the reply information structure, for example, the user interaction data is used to determine whether the first reply content and the second reply content are included in the reply information; the first reply content is the essential content of the reply information, and the first reply content can be understood as the answer of the user interaction data.

[0147] The reply information structure is used to determine whether the second reply content is included in the generated reply information. For example, the reply information structure includes the first reply content, or the first reply content and the second reply content. Whether the reply information includes the second reply content corresponds to the reply information structure, which can be understood as: the reply information structure includes the first reply content and the second reply content, and the corresponding reply information includes the second reply content; the reply information structure includes the first reply content, and the reply information does not include the second reply content.

[0148] In an optional embodiment, the user interaction data and the reply information structure are used to determine the essential content of the reply information, for example, the user interaction data and the reply information structure are used to determine the answer of the user interaction data.

[0149] In an optional embodiment, generating the reply information according to the user interaction data and the reply information structure can be: inputting the user interaction data and the reply information structure into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data and the reply information structure can be filled into a prompt template corresponding to the reply information structure to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0150] For example, generating the reply information according to the user interaction data and the reply information structure can be: generating reply indication information according to the user interaction data and the reply information structure, and generating the reply information according to the reply indication information and the user interaction data. Specifically, the user interaction data can be input into a first large language model to obtain the reply information structure output by the first large language model, and the reply indication information can be queried or generated according to the reply information structure. Input data can be determined according to the user interaction data and the reply indication information. The input data can be input into a second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0151] In an optional embodiment, generating the reply information according to the user interaction data, the target information amount, and the reply information structure can be: inputting the user interaction data, the target information amount, and the reply information structure into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data and the reply information structure can be filled into a prompt template corresponding to the target information amount and the reply information structure to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0152] For example, generating the reply information according to the user interaction data, the target information amount, and the reply information structure can be: generating reply indication information according to the user interaction data, the target information amount, and the reply information structure, and generating the reply information according to the reply indication information and the user interaction data. Specifically, the user interaction data can be input into a first large language model to obtain the target information amount and the reply information structure output by the first large language model, and the target information amount and the reply information structure can be used as the reply indication information. Input data can be determined according to the user interaction data and the reply indication information. The input data can be input into a second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0153] In an optional embodiment, the generated reply information is generally text data, which can be sent to the human-computer interaction device to display the reply information in text form to the user, or the generated reply information is an image displayed to the user, or the generated reply information is corresponding voice played to the user, or the generated reply information is corresponding video played to the user, etc. Alternatively, at least one data such as an image, voice, or video is generated according to the reply information, and the at least one data such as the image, voice, or video is sent to the human-computer interaction device.

[0154] In an example, the user's question content is: why did dinosaurs disappear. For example, the reply information structure includes a first reply content, which can be: wow, this is a very good question! Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky and hitting the earth with a loud noise, which is a big meteorite. The big stone caused a lot of dust and smoke to block the sky, and the sun's light could not reach the ground, and the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0155] In an example, the reply information structure includes a first reply content and a second reply content. The corresponding reply information includes: the main reason for the disappearance of dinosaurs is that a small planet hits the earth, causing the environment to change dramatically (first reply content). Do you want to know the story of the big meteorite hitting the earth (second reply content)?

[0156] For example, according to the user interaction data, the reply information can be generated as follows: according to the user interaction data, the reply indication information and the reply information structure are generated, and the reply information is generated according to the reply indication information, the reply information structure, and the user interaction data.

[0157] In an optional embodiment, the reply indication information is used to determine the information amount of the reply information, and the reply indication information, the reply information structure, and the user interaction data are used to cooperate to generate the reply information that meets the information amount corresponding to the reply indication information.

[0158] The embodiments of the present application can flexibly adjust the reply information structure of the reply information by determining the reply information structure matched with the user interaction data and generating the reply information corresponding to the user interaction data according to the reply information structure, thereby effectively improving the human-computer interaction experience.

[0159] Optionally, the generating of the reply information according to the user interaction data includes: determining a guide type corresponding to the user interaction data according to the user interaction data; and generating the reply information according to the user interaction data and the guide type, whether the reply information includes the second reply content corresponding to the guide type.

[0160] The guidance type is used to determine whether to generate the second reply content. Exemplarily, the guidance type includes a guided type and an unguided type. Whether the reply information includes the second reply content corresponds to the guidance type can mean that the guidance type is the guided type, and the reply information includes the second reply content; and the guidance type is the unguided type, and the reply information does not include the second reply content.

[0161] In an optional embodiment, the user interaction data includes current user interaction data and historical user interaction data; and the determining, according to the user interaction data, of the guidance type corresponding to the user interaction data includes determining, according to the current user interaction data and the historical user interaction data, of the guidance type corresponding to the user interaction data.

[0162] In an optional embodiment, the generating, according to the user interaction data and the guidance type, of the reply information can be inputting the user interaction data and the guidance type into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data and the guidance type can be filled into a prompt template corresponding to the guidance type to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0163] For another example, the generating, according to the user interaction data and the guidance type, of the reply information can be generating, according to the user interaction data and the guidance type, of reply indication information, and generating, according to the reply indication information and the user interaction data, of the reply information. Specifically, the user interaction data can be input into a first large language model to obtain the guidance type output by the first large language model, and the reply indication information can be queried or generated according to the guidance type. Input data can be determined according to the user interaction data and the reply indication information. The input data can be input into a second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0164] In an optional embodiment, the generating, according to the user interaction data, of the reply information includes determining, according to the user interaction data, of a target information amount corresponding to the user interaction data; determining, according to the user interaction data, of a guidance type corresponding to the user interaction data; and generating, according to the user interaction data, the target information amount, and the guidance type, of the reply information.

[0165] In an optional embodiment, according to the user interaction data, the target information quantity, and the guidance type, generating the reply information can be: inputting the user interaction data, the target information quantity, and the guidance type into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data and the guidance type can be filled into a prompt template corresponding to the target information quantity and the guidance type to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0166] For example, according to the user interaction data, the target information quantity, and the guidance type, generating the reply information can be: generating reply indication information according to the user interaction data, the target information quantity, and the guidance type, and generating the reply information according to the reply indication information and the user interaction data. Specifically, the user interaction data can be input into a first large language model to obtain the target information quantity and the guidance type output by the first large language model, and the target information quantity and the guidance type can be used as the reply indication information. Input data can be determined according to the user interaction data and the reply indication information. The input data can be input into a second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. The first large language model can also be replaced by a classification model.

[0167] According to the embodiments of the present application, the guidance type is obtained by processing the user interaction data, and the reply information including or not including the second reply content is generated according to the guidance type and the user interaction data. The guidance type can be directly extracted from the user interaction data, and the reply information can be generated based on the guidance type and the user interaction data. Whether the second reply content is generated can be accurately controlled.

[0168] In an optional embodiment, when the user interaction data is first user interaction data, a first guidance type corresponding to the first user interaction data is determined according to the first user interaction data. When the user interaction data is second user interaction data, a second guidance type corresponding to the second user interaction data is determined according to the second user interaction data. The user interaction data includes user attribute information. The guidance type corresponding to the user interaction data is determined according to the user attribute information. When the user attribute information is first user attribute information, a first guidance type corresponding to the first user attribute information is determined according to the first user attribute information. When the user attribute information is second user attribute information, a second guidance type corresponding to the second user attribute information is determined according to the second user attribute information. The first user attribute information is different from the second user attribute information, and the first guidance type is different from the second guidance type.

[0169] In an optional embodiment, the user attribute information is also used to indicate whether to generate the second reply content. The understanding ability of the user can be determined according to the user attribute information, and whether the second reply content adapted to the understanding ability of the user needs to be generated is determined according to the understanding ability of the user. In addition, the content of interest of the user is determined according to the user attribute information, and whether the second reply content adapted to the content of interest of the user needs to be generated is determined according to the content of interest of the user. The guidance type is different for different users with different user attribute information. For example, a mapping relationship between the user attribute information and the guidance type can be preset, and the corresponding guidance type is determined according to the user attribute information.

[0170] For example, the user attribute information includes a professional athlete, and the user question content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in the art work, and the second reply content can not be generated. For another example, the user attribute information includes a professional e-sports player, and the user question content is an introduction to XX game. It is determined that the user has a high interest in the game, and the second reply content can be generated. For another example, the user attribute information includes a user age of a child age. It is determined that the user has a weak understanding ability, and the second reply content can be generated. By determining whether to generate the second reply content based on the user attribute information in the user interaction data, the understanding ability and preferred content of the user can be considered to determine whether to add the second reply content to adapt to the understanding ability and preferred content of the user, and to assist the user to obtain more interesting and understandable content.

[0171] Optionally, the user interaction data includes: user question content; and the determining, according to the user interaction data, of the guidance type corresponding to the user interaction data includes: when a difficulty level corresponding to the user question content is a first difficulty level, determining, according to the user question content, a third guidance type corresponding to the user interaction data; when the difficulty level corresponding to the user question content is a second difficulty level, determining, according to the user question content, a fourth guidance type corresponding to the user interaction data; and the third guidance type is different from the fourth guidance type when the first difficulty level is different from the second difficulty level.

[0172] In an optional embodiment, difficult user question content can increase the second reply content, implement feedback by using more rounds of first reply content, facilitate the user to understand detailed and complete reply content, and thus determine that the guidance type is a guided type; simple user question content can not increase the second reply content, implement feedback by using less rounds of first reply content, facilitate the user to quickly understand the first reply content, and thus determine that the guidance type is a non-guided type. The guidance type is different for different difficulty levels of user question content. For example, a mapping relationship between the difficulty level and the guidance type can be preset, and the corresponding guidance type is determined according to the difficulty level.

[0173] By specifically defining the user interaction data as the user question content, whether to add the second reply content can be determined according to the difficulty of the user question content, the difficulty of the question content can be adapted to increase appropriate guide content, and the user can be assisted to obtain more detailed and complete reply content.

[0174] Optionally, the user interaction data includes: user historical dialogue; and the determining, according to the user interaction data, of the guide type corresponding to the user interaction data includes: when known information corresponding to the user historical dialogue is first known information, determining, according to the user historical dialogue, a fifth guide type corresponding to the user interaction data; when the known information corresponding to the user historical dialogue is second known information, determining, according to the user historical dialogue, a sixth guide type corresponding to the user interaction data; and when the first known information is different from the second known information, the fifth guide type is different from the sixth guide type.

[0175] In an optional embodiment, the user can select whether to repeatedly obtain the known information, and accordingly, the second reply content corresponding to the known information can be added, so as to determine the guide type as the guided type, and guide the user to obtain the reply content that has been replied but still wants to obtain. Alternatively, whether the user has ever involved in the content of the same field as the user question content can be determined according to the known information, so as to determine whether the user is interested in the content of the field, and when it is determined that the user is interested in the content of the field, the second reply content of the content of the field can be added, so as to determine the guide type as the guided type. The guide types are different for different known information historical dialogues. For example, a mapping relationship between the similarity of the known information and the user question content and the guide type can be preset, the similarity between the known information and the user question content is calculated, and the corresponding guide type is determined according to the similarity.

[0176] By specifically defining the user interaction data as the user historical dialogue, the known information can be considered to determine whether to add the second reply content, and the user can be assisted to obtain the content that has been obtained or the content that is interested in.

[0177] Optionally, the user interaction data includes: user behavior characteristics; and the determining, according to the user interaction data, of the guide type corresponding to the user interaction data includes: when preferred content corresponding to the user behavior characteristics is first preferred content, determining, according to the user behavior characteristics, a seventh guide type corresponding to the user interaction data; when the preferred content corresponding to the user behavior characteristics is second preferred content, determining, according to the user behavior characteristics, an eighth guide type corresponding to the user interaction data; and when the first preferred content is different from the second preferred content, the seventh guide type is different from the eighth guide type.

[0178] In an optional embodiment, the user wants to obtain more content of interest. Whether to add the second reply content can be determined according to the preference content, so as to determine the guidance type. For example, when the preference content is similar to the user query content, the guidance type is the guided type; when the preference content is not similar to the user query content, the guidance type is the unguided type. The guidance type is different for different preference content. For example, a mapping relationship between the similarity of the preference content and the user query content and the guidance type can be preset, the similarity between the preference content and the user query content is calculated, and the corresponding guidance type is determined according to the similarity.

[0179] By limiting the user interaction data to the user behavior feature, determining the preference content according to the user behavior feature, and determining the guidance type, more reply content can be provided for the user's interested question.

[0180] Optionally, the user interaction data includes vehicle data; and the determining the guidance type corresponding to the user interaction data according to the user interaction data includes: when the operation complexity corresponding to the vehicle data is a first operation complexity, determining a ninth guidance type corresponding to the user interaction data according to the vehicle data; when the operation complexity corresponding to the vehicle data is a second operation complexity, determining a tenth guidance type corresponding to the user interaction data according to the vehicle data; and the ninth guidance type is different from the tenth guidance type when the first operation complexity is different from the second operation complexity.

[0181] Generally, the more vehicle data needed by the user query content, the more complex the process of obtaining the first reply content, and the more reply content provided, which can be provided through multiple rounds of dialogue, that is, adding the second reply content, and the guidance type is the guided type; the less vehicle data needed by the user query content, the simpler the process of obtaining the first reply content, and there is no need to provide more content through multiple rounds of dialogue, that is, no second reply content is added, and the guidance type is the unguided type. Generally, the operation complexity is high, and the corresponding guidance type is the guided type; the operation complexity is low, and the corresponding guidance type is the unguided type.

[0182] The guidance type is different for different vehicle data. For example, a mapping relationship between the operation complexity and the guidance type can be preset, and the corresponding guidance type is determined according to the operation complexity.

[0183] By specifically limiting the user interaction data to the vehicle data, whether to add the second reply content can be considered according to the processing complexity of the vehicle data in the human-vehicle interaction scene, the guidance type can be adapted to the processing complexity of the vehicle data, and the user can be assisted to understand the state of the vehicle more comprehensively.

[0184] When the user interaction data includes at least two items of content, the guide type corresponding to the user interaction data can be determined according to the user interaction data as follows: for each item of content, a guide type corresponding to the item is determined; the guide types determined for each item of content are weighted to obtain a weighted result; and the weighted result is taken as the guide type corresponding to the user interaction data.

[0185] In one example, the user interaction data of the first user includes at least one of the first user attribute information, the first user question content of the first difficulty level, the first known information of the user history dialogue, the first preference content of the user behavior feature, and the first operation complexity of the vehicle data. At least one of the first guide type, the third guide type, the fifth guide type, the seventh guide type, and the ninth guide type can be weighted to obtain the guide type corresponding to the user interaction data.

[0186] In addition, the same user question content has different second reply contents corresponding to the user interaction data of different combinations of content.

[0187] In an optional embodiment, the reply information is generated according to the user interaction data, including: determining the target information amount and the guide type according to the user interaction data, and determining the reply type; querying the reply indication information corresponding to the reply type from the plurality of preset reply indication information; and generating the reply information corresponding to the user interaction data according to the reply indication information and the user interaction data.

[0188] In an optional embodiment, the reply type is used to query the reply indication information. The reply indication information can include types such as little information amount and no guide follow-up question, little information amount and guide follow-up question, much information amount and no guide follow-up question, and much information amount and guide follow-up question. The amount of information can be determined according to the target information amount, and the guide can be determined according to the guide type, so as to obtain the reply type by combining the target information amount and the guide type.

[0189] The plurality of reply indication information can be pre-configured, and a correspondence between the reply indication information and the reply type is established, and according to the reply type, the reply indication information corresponding to the reply type is queried in the correspondence. For example, the reply indication information corresponding to the reply type with less information and without guided follow-up questions can be: please provide a simple answer to the question: XX. For another example, the reply indication information corresponding to the reply type with less information and with guided follow-up questions can be: please provide an answer that a child can understand to the question: XX, and add guided content that meets the child's interests. For another example, the reply indication information corresponding to the reply type with more information and without guided follow-up questions can be: please provide a complete answer to the question: XX. For another example, the reply indication information corresponding to the reply type with more information and with guided follow-up questions can be: please provide a detailed answer to the question: XX, and add guided content for more answers.

[0190] By configuring the reply type and the reply indication information corresponding to each reply type, different reply requirements can be preset, and the corresponding reply indication information can be provided according to the reply requirements, and the reply content that best matches the reply requirements can be generated based on the reply indication information, so that the reply requirements can be adapted, accurate reply content can be provided, and the flexibility of the reply content can be increased to meet various interactive requirements and adapt to various interactive scenarios. Meanwhile, the reply indication information is preset, and the correspondence between the reply type and the reply indication information is established, so that the reply indication information can be quickly determined, thereby improving the interactive efficiency.

[0191] In an optional embodiment, the reply information corresponding to the type with less information and without guided follow-up questions only includes the first reply content and does not include the second reply content, and the information amount of the first reply content is relatively small; the reply information corresponding to the type with less information and with guided follow-up questions includes the first reply content and the second reply content, and the information amount of the first reply content is relatively small; the reply information corresponding to the type with more information and without guided follow-up questions only includes the first reply content and does not include the second reply content, and the information amount of the first reply content is relatively large; and the reply information corresponding to the type with more information and with guided follow-up questions includes the first reply content and the second reply content, and the information amount of the first reply content is relatively large.

[0192] As in the foregoing example, the reply information corresponding to the type with less information and without guided follow-up questions includes: the main reason for the disappearance of dinosaurs is that an asteroid hits the earth, causing environmental changes (first reply content).

[0193] For another example, the reply information corresponding to the type with less information and with guided follow-up questions includes: the main reason for the disappearance of dinosaurs is that an asteroid hits the earth, causing environmental changes (first reply content). Do you want to know the story of this large meteorite hitting the earth (second reply content)?

[0194] For another example, the reply information corresponding to the type of information quantity being large and no guided follow-up question includes: Wow, this is a very good question! Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky and hitting the earth with a loud noise, which is a large meteorite. This large stone caused a lot of dust and smoke to cover the sky, and the sun's light could not reach the ground, and the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also gradually disappeared (first reply content).

[0195] For another example, the reply information corresponding to the type of information quantity being large and no guided follow-up question includes: Wow, this is a very good question! Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky and hitting the earth with a loud noise, which is a large meteorite. This large stone caused a lot of dust and smoke to cover the sky, and the sun's light could not reach the ground, and the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also gradually disappeared (first reply content).

[0196] By configuring multiple reply types and corresponding reply indication information for each reply type, the diversity of reply types is increased, the richness of reply indication information is increased, and various interactive scenarios can be adapted and the information quantity of reply content can be flexibly adjusted.

[0197] Optionally, the generating of the reply indication information according to the user interaction data includes: inputting the user interaction data into a classification model to obtain a reply type output by the classification model; and querying, according to the reply type, the reply indication information corresponding to the reply type in a preset reply prompt template.

[0198] In an optional embodiment, the classification model is used to classify the user interaction data to obtain a reply type. For example, the classification model can be a large language model. The specific processing process of the classification model is: encoding the user interaction data to obtain a corresponding vector representation, and decoding the vector to obtain a reply type.

[0199] By inputting the user interaction data into the classification model, the reply type can be obtained, complex user interaction data can be processed, and reply information can be quickly generated.

[0200] Optionally, the reply indication information includes a reply prompt template; and the generating of the reply information corresponding to the user interaction data according to the reply indication information and the user interaction data includes: filling the user interaction data into the reply prompt template to obtain target input content; and inputting the target input content into a generation model to obtain the reply information corresponding to the user interaction data.

[0201] In an optional embodiment, the reply prompt template comprises a slot of user interaction data, and filling the user interaction data into the reply prompt template is adding the user interaction data into the corresponding slot of the reply prompt template, and the filled reply prompt template is the target input content. The target input content is used as input data of the generation model. The generation model is used to process the input target input content to generate reply information. For example, the specific processing process of the generation model is: encoding the target input content to obtain a corresponding vector representation, decoding the vector representation to obtain the reply information. The generation model can be a large language model used to input natural language content to generate reply information.

[0202] By limiting the reply indication information to the reply prompt template, inputting the user interaction data into the reply prompt template to obtain the target input content, and inputting the target input content into the generation model to obtain the reply information, complex question content can be processed to generate accurate and coherent reply information, improve the authenticity of interaction, and quickly generate reply information to improve the efficiency of interaction.

[0203] It should be noted that the aforementioned classification model and generation model can be the same model or two different independent models. Optionally, the user interaction data is input into the classification model to obtain the reply type output by the classification model; the reply indication information corresponding to the reply type is queried, and the user interaction data is filled into the reply indication information to obtain the target input content; and the target input content is input into the generation model to obtain the reply information output by the generation model. The size of the classification model is smaller than the size of the generation model, specifically, the number of layers of the classification model is smaller than the number of layers of the generation model, and / or the number of parameters of the classification model is smaller than the number of parameters of the generation model. In this way, the user interaction data is processed by the two-level connected model structure, the simple operation of generating the reply type can be performed by the pre-positioned small-scale and fast classification model, and the complex operation of generating the reply information can be performed by the post-positioned large-scale, powerful reasoning and production capacity, and better instruction following ability generation model, which can reasonably allocate tasks, improve the generation speed of the reply information on the basis of considering the accuracy of the reply information, and achieve the balance between cost and effect.

[0204] Optionally, the generation model is obtained by: obtaining standard question content and standard reply information corresponding to the standard question content; splitting the standard reply information to obtain a plurality of sub-reply information and a reply order of each of the sub-reply information; the information amount of the standard reply information is greater than the information amount of each of the sub-reply information; adding corresponding guide follow-up question content to the sub-reply information with the first reply order to obtain target reply information; combining at least one of the target reply information and the sub-reply information with the last reply order with the standard question content to generate a plurality of training samples; and fine-tuning a pre-trained initial model using each of the training samples to obtain the generation model.

[0205] In an optional embodiment, the initial model can be a pre-trained large language model adapted to a general field. The generation model can be a model obtained by fine-tuning the large language model adapted to the general field to adapt to the reply content generation method provided in the embodiments of the present application. The standard question content and the standard reply information are a pair of question and reply data, and the standard reply information is the correct reply information of the standard question content. It should be noted that there can be more than one correct reply information of the standard question content, and one correct reply information can be selected as the standard reply information, or a plurality of correct reply information can be selected to be fused to obtain the standard reply information. The way of obtaining the standard question content and the standard reply information can include at least one of the following: artificial generation, obtaining high-score question and answer pairs and theme content and corresponding comments in a question and answer website, etc.

[0206] In an optional embodiment, the splitting of the standard reply information can be splitting the standard reply information into sub-reply information with a total score logical relationship, or can be splitting the standard reply content into sub-reply information with a progressive relationship or an inference relationship. The information amount of the standard reply information is greater than the information amount of the sub-reply information.

[0207] For example, the standard reply information is: Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky and hitting the earth with a loud noise, which is a large meteorite. The large stone caused a lot of dust and smoke to block the sky, and the sun's light could not reach the ground, so the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also gradually disappeared.

[0208] In an optional embodiment, the total sub-reply information in the split sub-reply information is: the disappearance of dinosaurs is because of the asteroid impact on the earth, which causes the drastic change of the environment, leading to the inability to survive. One of the sub-reply information is: a huge stone comes from the sky and crashes onto the earth with a loud noise, which is a large meteorite. The large stone causes a lot of dust and smoke to cover the sky, making the earth very cold, leading to the disappearance of dinosaurs. One of the sub-reply information is: the earth becomes very cold, and many plants die. Dinosaurs have no food to eat, so they also slowly disappear.

[0209] For example, the standard reply information is: S1 / T1=V1, S2 / T2=V2, V1>V2. One of the sub-reply information in the split sub-reply information is: S1 / T1=V1; one of the sub-reply information in the split sub-reply information is: S2 / T2=V2; one of the sub-reply information in the split sub-reply information is: V1>V2.

[0210] In the foregoing example, each sub-reply information only includes part of the content in the standard reply information, and thus the information amount of the standard reply information is greater than the information amount of each sub-reply information.

[0211] In addition, there are other split manners and relationships between sub-reply information, which can be set as needed and are not specifically limited.

[0212] The reply order can indicate the output order of each sub-reply information obtained by splitting the same standard reply information. The reply order of each sub-reply information can be determined according to the logical relationship between the sub-reply information.

[0213] For example, for the sub-reply information of the total and sub logical relationship, the reply order of the total sub-reply information is usually in front, and the reply order of the sub sub-reply information is in the back. In addition, each sub sub-reply information can be determined according to the importance, causal relationship or other logical relationship. If each sub sub-reply information is in a parallel relationship, the reply order of each sub-reply information can be randomly determined.

[0214] For example, the sub-reply information for the progressive relationship or the reasoning relationship can be determined according to the progressive relationship and the reasoning relationship. Specifically, in the reasoning relationship, the reply order of the sub-reply information without dependency is in the front, and the reply order of the sub-reply information dependent on other sub-reply information is after the reply order of the other sub-reply information on which it depends.

[0215] In an optional embodiment, the sub-reply information is divided into two categories, one is the sub-reply information at the end of the reply sequence, and the other is the sub-reply information other than the sub-reply information at the end of the reply sequence, i.e., the sub-reply information at the beginning of the reply sequence. The number of the sub-reply information at the beginning of the reply sequence is at least one, and the sub-reply information at the beginning of the reply sequence is the sub-reply information other than the sub-reply information at the end of the reply sequence. The sub-reply information at the beginning of the reply sequence can add a leading follow-up content to obtain the target reply information. After the sub-reply information at the beginning of the reply sequence adds the leading follow-up content, it can guide the user to further follow up the sub-reply information at the end of the reply sequence. The leading follow-up content can be determined according to the standard question content, or a unified language can be used, or it can be determined according to the keywords or the content summary of the first sentence of the next sub-reply information adjacent to the sub-reply information in the reply sequence. It can be set according to the needs, and it is not specifically limited. The sub-reply information at the end of the reply sequence does not have sub-reply information later in the reply sequence, so it does not need to add a leading follow-up content in the sub-reply information at the end of the reply sequence.

[0216] In an optional embodiment, for the same standard question content, each target reply information is combined with the standard question content to obtain a corresponding training sample. The sub-reply information at the end of the reply sequence is combined with the standard question content to obtain a corresponding training sample. The number of sub-reply information obtained by splitting the standard reply information is the same as the number of generated training samples.

[0217] In addition, a large number of standard question contents and corresponding standard question content pairs can also be set, and for each pair, a plurality of training samples are generated. A large number of training samples are used to fine-tune the initial model. For example, the initial model can be fine-tuned in an SFT (Self-optimizing and Self-adjusting) manner.

[0218] In addition to being able to split the standard reply information to obtain sub-reply information, the standard reply information and the standard question content can be combined to generate training samples, and at least one simplified manner can be used to simplify the standard reply information to obtain at least one sub-reply information, each of which is combined with the standard question content to generate a corresponding training sample. In addition, at least one associated information of the standard reply information can be obtained, and the standard reply information can be expanded to obtain at least one reply information with more information, and each expanded reply information is combined with the standard question content to generate a corresponding training sample. The above generated training samples can be input into the initial model for fine-tuning. This greatly expands the coverage of the training samples, improves the representativeness of the training samples, and fine-tunes the initial model using these training samples, which can improve the generalization ability of the fine-tuned generation model, improve the accuracy of the reply content of the generation model, and increase the diversity of the reply content.

[0219] By fine-tuning the initial model, the generation model can distinguish the information amount of the user interaction data adjusted reply information and guide the content production, and can gradually guide the user interaction, gradually improve the dialogue authenticity.

[0220] In a specific application scenario, as shown in FIG. 4, in the vehicle interaction process, the vehicle obtains the user question content and the multi-modal data and sends them to the server. The server executes the reply information generation method as follows:

[0221] S401, obtaining user interaction data.

[0222] In an optional embodiment, the user interaction data at least includes user question content and user attribute information. The user interaction data can also include, but is not limited to, user historical dialogue, user behavior characteristics, and vehicle data, etc. The user attribute information at least includes user identification and age. The user interaction data can be multi-modal data and can include data of at least one media type.

[0223] S402, the server outputs the determination result of adult or child through the classification model.

[0224] In an optional embodiment, the target information amount corresponding to the adult is large, and the target information amount corresponding to the child is small.

[0225] S403, when determining the adult dialogue link, the server outputs the determination result of whether there is a dialogue guide phrase through the classification model.

[0226] In an optional embodiment, no dialogue guide phrase corresponds to a guide type of no guide type. There is a dialogue guide phrase, which corresponds to a guide type of guided type.

[0227] S404, in the case of no dialogue guidance, an adult and no guidance corresponding prompt template is obtained.

[0228] S405, in the case of dialogue guidance, an adult and guided corresponding prompt template is obtained.

[0229] S406, in the case of a child dialogue link, the server outputs the judgment result of whether there is dialogue guidance through the classification model.

[0230] S407, in the case of no dialogue guidance, a child and no guidance corresponding prompt template is obtained.

[0231] S408, in the case of dialogue guidance, a child and guided corresponding prompt template is obtained.

[0232] In an optional embodiment, according to the child or adult, the target information amount is determined, according to the dialogue guidance or not, the guidance type is determined, according to the target information amount and the guidance type, the reply type is determined, and the reply type corresponding reply indication information is queried, and the reply indication information can be a prompt template.

[0233] In an optional embodiment, through the classification model, according to the user age, the dialogue context (i.e. historical dialogue) and the user question content, the child or adult (information amount) is judged and the guidance or not is judged, that is, the reply type is determined, and the prompt template prompt input to the generation model is selected. Among them, the child corresponds to the small amount of information in the foregoing, and the adult corresponds to the large amount of information in the foregoing.

[0234] In an optional embodiment, the classification model first determines whether the adult dialogue link or the child dialogue link according to the age, the adult dialogue link corresponds to the reply type with large amount of information, and the child dialogue link corresponds to the reply type with small amount of information. Secondly, the classification model judges the guidance or not of the reply type according to the user question content and the user information. Or the classification model judges the guidance or not of the reply type according to the user question content, the user information and the vehicle information.

[0235] S409, the user question content, the multi-modal data and the prompt template are fused.

[0236] S410, the fusion result is input into the model, and the model outputs the reply content and the dialogue guidance.

[0237] In an optional embodiment, the dialogue guidance can be empty. The fusion result is input into the generation model, and the reply information output by the generation model is obtained, and the reply information at least includes the reply content (first reply content), and optionally includes the guidance follow-up question content (second reply content).

[0238] The classification model is a small-sized pre-model, and the generation model is a large-sized post-model. The classification model is also a large language model, and the classification model can perform some simple tasks. Simple tasks, such as vehicle control tasks, can be directly input into the classification model, the classification model generates vehicle control instructions, and the vehicle control instructions are sent to the vehicle machine, and the vehicle machine controls some entity modules of the vehicle, without the need for processing in the input into the generation model. For example, the entity module is a vehicle window, and the vehicle control instruction is an instruction for controlling the opening and closing of the vehicle window. For another example, the classification model can also call a search interface according to the user query content, obtain external knowledge related to the user query content, send the external knowledge of the user query content to the large language model, specifically fill the external knowledge into the reply instruction information, realize the fusion of the user query content and the external knowledge, obtain the input data of the large language model, so that the large language model obtains the external knowledge and better understands the user query content, thereby improving the reply accuracy. In this way, the high-configured large language model can be reduced to implement all tasks, thereby causing the cost of task execution to be relatively high, and through the two-level model structure of low configuration and high configuration, the tasks can be reasonably distributed, and the cost and the accuracy of the reply content can be considered.

[0239] The embodiment of the application can solve the problem of missing multi-modal information in the previous model architecture by fusing multi-modal data combined with user information, can perceive more and richer information, and thus improve the accuracy of the output reply content; by appropriately splitting the dialogue process, gradually guiding the user, sequentially and progressively, simulating the dialogue between users, the user query can be guided, the authenticity of the dialogue reply content can be improved, and the user experience can be improved; at the same time, the reply content can be shortened, the user can understand the reply content, and the human-computer interaction experience of children is improved; through reasonable multi-level model distribution, the cost and effect balance can be achieved, and thus the cost and response time delay can be reduced.

[0240] FIG. 5 is a flowchart of another reply information generation method according to an embodiment of the application.

[0241] In an optional embodiment, the step of determining the reply information structure according to the user interaction data is refined as: determining the reply information structure according to the current user interaction data and / or the historical user interaction data.

[0242] It should be noted that the parts not described in detail in the embodiments of the application can be referred to the descriptions of the foregoing embodiments.

[0243] Referring to the reply information generation method shown in FIG. 5, the method comprises the following steps.

[0244] S501, obtaining user interaction data.

[0245] S502, determining the reply information structure according to the current user interaction data and / or the historical user interaction data.

[0246] In addition, the user interaction data of the same user question content in different combinations has different corresponding reply information structures.

[0247] S503, generating reply information according to the user interaction data and the reply information structure.

[0248] The embodiment increases the flexibility of the reply information structure by flexibly combining the content of the user interaction data, so as to meet various interaction requirements and adapt to various interaction scenarios.

[0249] FIG. 6 is a flowchart of another reply information generation method according to an embodiment of the present application.

[0250] In an optional embodiment, the step of "determining a reply information structure according to the user interaction data" is further detailed as follows: when the user interaction data is first user interaction data, determining, according to the first user interaction data, that a reply information structure corresponding to the first user interaction data is a first reply information structure; and when the user interaction data is second user interaction data, determining, according to the second user interaction data, that a reply information structure corresponding to the second user interaction data is a second reply information structure.

[0251] It should be noted that the parts not described in detail in the embodiments of the present application can be referred to the descriptions of the foregoing embodiments.

[0252] Referring to the reply information generation method shown in FIG. 6, the method comprises the following steps.

[0253] S601, obtaining user interaction data.

[0254] S602, when the user interaction data is first user interaction data, determining, according to the first user interaction data, that a reply information structure corresponding to the first user interaction data is a first reply information structure.

[0255] The first reply information structure refers to a structure including first reply content.

[0256] In an example, the first user interaction data can include, but is not limited to, at least one of the following: first user attribute information; user question content with a difficulty level of a first difficulty level; user historical dialogue with known information of a first known information; user behavior characteristics with preferred content of a first preferred content; and vehicle data with an operation complexity of a first operation complexity.

[0257] In an optional embodiment, the step of "determining, according to the first user interaction data, that a reply information structure corresponding to the first user interaction data is a first reply information structure" comprises at least one of the following:

[0258] In a case where the first user interaction data is first user attribute information, a reply information structure corresponding to the first user attribute information is determined according to the first user attribute information, and the reply information structure is determined to contain first reply content.

[0259] In a case where the user question content corresponds to a first difficulty level, a reply information structure corresponding to the user question content is determined according to the user question content, and the reply information structure is determined to contain first reply content.

[0260] In a case where known information corresponding to the user historical conversation contained in the user historical interaction data is first known information, a reply information structure corresponding to the user interaction data is determined according to the user historical conversation, and the reply information structure is determined to contain first reply content.

[0261] In a case where the user behavior feature corresponds to first preference content, a reply information structure corresponding to the user interaction data is determined according to the user behavior feature, and the reply information structure is determined to contain first reply content.

[0262] In a case where the vehicle data corresponds to a first operation complexity level, a reply information structure corresponding to the user interaction data is determined according to the vehicle data, and the reply information structure is determined to contain first reply content.

[0263] S603, in a case where the user interaction data is second user interaction data, a reply information structure corresponding to the second user interaction data is determined according to the second user interaction data, and the reply information structure is determined to be second reply information structure.

[0264] In an example, the second user interaction data can include, but is not limited to, at least one of the following: second user attribute information; user question content with a second difficulty level; user historical conversation with second known information; user behavior feature with second preference content; and vehicle data with a second operation complexity level.

[0265] In an optional embodiment, according to the second user interaction data, the reply information structure corresponding to the second user interaction data is determined as a second reply information structure, which includes at least one of the following: when the second user interaction data is second user attribute information, according to the second user attribute information, the reply information structure corresponding to the first user attribute information is determined as a structure containing first reply content and second reply content; when the difficulty level corresponding to the user query content is a second difficulty level, according to the user query content, the reply information structure corresponding to the user query content is determined as a structure containing first reply content and second reply content; when the known information corresponding to the user historical conversation contained in the user historical interaction data is second known information, according to the user historical conversation, the reply information structure corresponding to the user interaction data is determined as a structure containing first reply content and second reply content; when the preference content corresponding to the user behavior feature is second preference content, according to the user behavior feature, the reply information structure corresponding to the user interaction data is determined as a structure containing first reply content and second reply content; and when the operation complexity corresponding to the vehicle data is a second operation complexity, according to the vehicle data, the reply information structure corresponding to the user interaction data is determined as a structure containing first reply content and second reply content.

[0266] In an optional embodiment, the user attribute information is also used to indicate whether to generate the second reply content. The understanding ability of the user can be determined according to the user attribute information, and whether the second reply content adapted to the understanding ability of the user needs to be generated is determined according to the understanding ability of the user. In addition, the interested content of the user is determined according to the user attribute information, and whether the second reply content adapted to the interested content of the user needs to be generated is determined according to the interested content of the user. The reply information structure is different for different users with different user attribute information. For example, a mapping relationship between the user attribute information and the reply information structure can be preset, and the corresponding reply information structure is determined according to the user attribute information.

[0267] For example, the user attribute information includes a profession of an athlete, and the user query content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in art works, and the second reply content can not be generated, i.e., the reply information structure contains only the first reply content.

[0268] For example, the user attribute information includes a profession of an athlete, and the user query content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in art works, and the second reply content can not be generated, i.e., the reply information structure contains only the first reply content.

[0269] For example, the user attribute information includes a profession of an athlete, and the user query content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in art works, and the second reply content can not be generated, i.e., the reply information structure contains only the first reply content.

[0269] For example, the user attribute information includes a profession of an athlete, and the user query content is the symbolic meaning of portrait XX. It is determined that the user has a low interest in art works, and the second reply content can not be generated, i.e., the reply information structure contains only the first reply content.

[0270] By determining whether to generate the second reply content based on the user attribute information in the user interaction data, the understanding ability and preferred content of the user can be considered, and the understanding ability and preferred content of the user are adapted to determine whether to add the second reply content, thereby assisting the user to obtain more interesting and understandable content.

[0271] In an optional embodiment, whether to generate the second reply content can be determined based on the difficulty level corresponding to the user question content. In a case where the difficulty level corresponding to the user question content is a first difficulty level, it is indicated that the user question content is not difficult to understand, and the first reply content as an answer can be directly replied in full, without adding the second reply content.

[0272] In an optional embodiment, whether to generate the second reply content can be determined based on the difficulty level corresponding to the user question content. In a case where the difficulty level corresponding to the user question content is a second difficulty level, it is indicated that the user question content is difficult to understand, and the first reply content as an answer can be partially replied, and the second reply content is added, so that the user quickly understands the answer of the first reply content.

[0273] In an optional embodiment, the difficult user question content can increase the second reply content, realize feedback by using more rounds of the first reply content, facilitate the user to understand the detailed and complete reply content, and thereby determine the reply information structure as containing the first reply content and the second reply content; the simple user question content can not increase the second reply content, realize feedback by using less rounds of the first reply content, facilitate the user to quickly understand the first reply content, and thereby determine the reply information structure as containing the first reply content. For user question contents of different difficulty levels, the guidance types are different. For example, a mapping relationship between the difficulty levels and the guidance types can be preset, and the corresponding reply information structure is determined according to the difficulty level.

[0274] By specifically limiting the user interaction data to the user question content, whether to add the second reply content can be determined according to the difficulty level of the user question content, and the difficulty level of the question content can be adapted to increase appropriate guidance content, thereby assisting the user to obtain more detailed and complete reply content.

[0275] In an optional embodiment, the user can select whether to repeat the acquisition of the known information, and accordingly, the second reply content corresponding to the known information can be added, so as to determine the reply information structure as containing the first reply content and the second reply content, and guide the user to acquire the reply content that has been replied to in the history but still wants to acquire. Alternatively, it can be determined according to the known information whether the user has inquired into the content in the same field as the user's inquiry content, so as to determine whether the user is interested in the content in the field, and when it is determined that the user is interested in the content in the field, the second reply content of the content in the field can be added, so as to determine the reply information structure as containing the first reply content and the second reply content. The reply information structure is different for different known information history dialogues. For example, a mapping relationship between the similarity of the known information and the user's inquiry content and the guide type can be preset, the similarity between the known information and the user's inquiry content is calculated, and according to the similarity, the corresponding reply information structure is determined.

[0276] By specifically limiting the user interaction data to the user history dialogue, the known information can be considered to determine whether to add the second reply content, which can assist the user to acquire the content that has been acquired in the history or the content that is interested in.

[0277] In an optional embodiment, the user wants to acquire more content that is interested in. Whether to add the second reply content can be determined according to the preference content, so as to determine the reply information structure. For example, when the preference content is similar to the user's inquiry content, the reply information structure is determined as containing the first reply content and the second reply content; when the preference content is not similar to the user's inquiry content, the reply information structure is determined as containing the first reply content. The reply information structure is different for different preference content. For example, a mapping relationship between the similarity of the preference content and the user's inquiry content and the guide type can be preset, the similarity between the preference content and the user's inquiry content is calculated, and according to the similarity, the corresponding reply information structure is determined.

[0278] By limiting the user interaction data to the user behavior feature, the preference content determined according to the user behavior feature, and the reply information structure, more reply content can be provided for the user's interested question.

[0279] Generally, the more vehicle data needed by the user question content, the more complex the process of obtaining the first reply content, and the more reply content provided, which can be provided through multiple rounds of dialogue, i.e., adding the second reply content, to determine the reply information structure as containing the first reply content and the second reply content; the less vehicle data needed by the user question content, the simpler the process of obtaining the first reply content, and no more content needs to be provided through multiple rounds of dialogue, i.e., no second reply content is added, to determine the reply information structure as containing the first reply content. Generally, the operation is complex, and the corresponding reply information structure is the one containing the first reply content and the second reply content; the operation is simple, and the corresponding reply information structure is the one containing the first reply content.

[0280] S604, generating reply information according to the user interaction data and the reply information structure.

[0281] In an optional embodiment, FIG. 7 is a flowchart of still another method for generating reply information according to an embodiment of the present application. This embodiment is based on the above-mentioned embodiments, and adds the determination of initial reply content based on user interaction data and the splitting of the initial reply content.

[0282] It should be noted that the parts not described in detail in the embodiments of the present application can refer to the descriptions of the above-mentioned embodiments.

[0283] Referring to the method for generating reply information shown in FIG. 7, the method comprises the following steps.

[0284] S701, obtaining user interaction data.

[0285] S702, determining a reply information structure according to the user interaction data.

[0286] S703, generating reply information according to the user interaction data and the reply information structure.

[0287] S704, determining initial reply content according to the user interaction data.

[0288] In an example, the initial reply content refers to the reply content of the complete answer determined according to the user interaction data. In an embodiment, the process of determining the initial reply content according to the user interaction data can refer to the process of determining the first reply content only according to the user interaction data, which is not described herein again.

[0289] S705, determining the splitting of the initial reply content according to the user interaction data.

[0290] In an example, the user interaction data can be user attribute information. For example, the user attribute information can include, but is not limited to, at least one of the following: user identification, gender, age, education level, occupation, and the like. The user identification is used to represent the unique identification of the user; the education level is used to represent the degree of knowledge mastered by the user, for example, the education level is classified according to the educational level, which can include, but is not limited to, at least one of the following: pre-school education, primary school, middle school, junior college, undergraduate, graduate, and the like; the education level is classified according to the professional skill education background, which can include, but is not limited to, at least one of the following: science and engineering major, liberal arts major, art and design major, and the like. The occupation is used to represent the type of work engaged in by the user, which is associated with factors such as personal skills, knowledge, work environment, and responsibilities; the occupation classification can include, but is not limited to, at least one of the following: student, engineer (such as software engineer, mechanical engineer, electrical engineer, and the like), doctor, teacher, and the like. In an optional embodiment, based on the user interaction data, it can be determined whether the initial reply content needs to be split to obtain multiple first reply contents and the reply order of each first reply content; wherein the information amount contained in each first reply content is less than the information amount contained in the initial reply content; and the final reply information each time is obtained by adding the corresponding guide follow-up question content to the first reply content with the earlier reply order.

[0291] In an optional embodiment, in the case of splitting the initial reply content, the initial reply content can be split into multiple first reply contents according to the total score logical relationship; or the initial reply content can be split into multiple first reply contents according to the progressive relationship; or the initial reply content can be split into multiple first reply contents according to the reasoning relationship.

[0292] In an optional embodiment, the initial reply content can include two types of first reply content, one is the first reply content with the last reply order, and the other is the first reply content other than the first reply content with the last reply order, i.e., the first reply content with the reply order other than the last. The number of the first reply content with the last reply order is one, while the number of the first reply content with the reply order other than the last can be one or more. In an optional embodiment, the guide follow-up question content, i.e., the second reply content, can be added after the first reply content with the reply order other than the last, which can guide the user to further ask the first reply content with the later reply order.

[0293] In an optional embodiment, the second reply content can be determined according to the specific content of the first reply content, can be determined by using a unified phrase, or can be determined according to the content of the keywords or the first sentence of the next first reply content adjacent in the reply order of the sub-reply information. It can be set as needed, and is not specifically limited. The last first reply content in the reply order does not exist the first reply content later in the reply order, and therefore, the leading follow-up content does not need to be added in the last first reply content in the reply order.

[0294] For example, if the user's education level is a bachelor's degree or above, and the user interaction data is a question related to the user's own occupation, the initial reply content can be directly replied to the user's question in one time without splitting, so as to reduce the process of the user's tedious questioning and greatly improve the user's experience; if the user's education level is pre-school education, and the user interaction data is a relatively complex question, the user's understanding ability is weak, and the initial reply content can be split to make the user obtain a smaller amount of information each time to facilitate the user's understanding.

[0295] In an optional embodiment, the "determining the splitting of the initial reply content according to the user interaction data" is refined as: determining the splitting of the initial reply content according to the current user interaction data and / or the historical user interaction data.

[0296] In an optional embodiment, the user interaction data includes: current user interaction data and historical user interaction data; and the determining the splitting of the initial reply content according to the user interaction data includes one of: determining the splitting of the initial reply content according to the user attribute information and the user identity information contained in the current user interaction data; determining the splitting of the initial reply content according to the difficulty level of the user questioning content and the user identity information contained in the current user interaction data; determining the splitting of the initial reply content according to the historical user interaction data and the user identity information; determining the splitting of the initial reply content according to the user behavior characteristics and the user identity information contained in the current user interaction data; and determining the splitting of the initial reply content according to the vehicle data and the user identity information contained in the current user interaction data.

[0297] In an optional embodiment, the splitting of the initial reply content can be determined based on the user attribute information and the user identity information contained in the current user interaction data. The user attribute information and the user identity information can also be used to indicate whether the initial reply content is split. The understanding ability and the professional ability of the user can be determined based on the user attribute information and the user identity information, and whether the initial reply content needs to be split can be determined according to the understanding ability and the professional ability.

[0298] For example, assuming that the user attribute information includes an age of 4 years old and the user question content is "why is the earth round?", it can be determined that the user's understanding ability and professional ability are low, at which time the initial reply content can be split into multiple first reply contents to facilitate the user's understanding. And the second reply content is added in each first reply content which is not the last in the reply sequence to guide the user to ask questions, which facilitates the user's understanding while satisfying the user's curiosity.

[0299] In an optional embodiment, the splitting of the first reply content can be determined according to historical user interaction data and user identity information. The historical user interaction data can include: user historical dialogue, which is used to determine whether the user has obtained content in the same field as the user's current question content, so as to determine whether the user is interested in the content in this field and the degree of understanding, to determine whether to split the initial reply content; if it is determined based on the user historical dialogue that the user only has less understanding of the content in the same field as the current question content, it is determined that the initial reply content is split, the user can be provided with multiple rounds of follow-up questions, and the amount of information of the sub-reply information provided each time is less to facilitate the user's understanding; if it is determined based on the user historical dialogue that the user has more understanding of the content in the same field as the current question content, it is determined that the initial reply content is not split, and detailed and complete reply content can be directly provided to the user to reduce multiple rounds of follow-up questions to facilitate the user to understand the initial reply content more quickly.

[0300] In an optional embodiment, determining the target information amount corresponding to the user interaction data according to the user interaction data comprises: when the user interaction data is first user interaction data, determining the target information amount corresponding to the first user interaction data as a first information amount according to the first user interaction data; when the user interaction data is second user interaction data, determining the target information amount corresponding to the second user interaction data as a second information amount according to the second user interaction data.

[0301] In an optional embodiment, "determining the splitting of the initial reply content according to the user interaction data" is refined as: when the user interaction data is first user interaction data, determining the splitting of the initial reply content according to the first user interaction data as splitting; when the user interaction data is second user interaction data, determining the splitting of the initial reply content according to the second user interaction data as not splitting.

[0302] In an optional embodiment, the splitting of the initial reply content is determined according to the user interaction data, including at least one of the following: when the first user interaction data is first user attribute information, the splitting of the initial reply content is determined to be split according to the first user attribute information; when a difficulty level corresponding to the user question content is a first difficulty level, the splitting of the initial reply content is determined to be split according to the user question content; when known information corresponding to a user historical conversation included in the user historical interaction data is first known information, the splitting of the initial reply content is determined to be split according to the user historical conversation; when preferred content corresponding to the user behavior feature is first preferred content, the splitting of the initial reply content is determined to be split according to the user behavior feature; when an operation complexity corresponding to the vehicle data is a first operation complexity, the splitting of the initial reply content is determined to be split according to the vehicle data.

[0303] In an optional embodiment, the second user interaction data is determined to correspond to a second reply information structure according to the second user interaction data, including at least one of the following: when the second user interaction data is second user attribute information, the splitting of the initial reply content is determined not to be split according to the second user attribute information; when a difficulty level corresponding to the user question content is a second difficulty level, the splitting of the initial reply content is determined not to be split according to the user question content; when known information corresponding to a user historical conversation included in the user historical interaction data is second known information, the splitting of the initial reply content is determined not to be split according to the user historical conversation; when preferred content corresponding to the user behavior feature is second preferred content, the splitting of the initial reply content is determined not to be split according to the user behavior feature; when an operation complexity corresponding to the vehicle data is a second operation complexity, the splitting of the initial reply content is determined not to be split according to the vehicle data.

[0304] In an optional embodiment, the splitting of the initial reply content can be determined according to the first user interaction data. When the difficulty level corresponding to the user query content is a first difficulty level, and the user's understanding ability and professional ability are determined to be good based on the user identity information, the initial reply content can not be split, and detailed and complete reply content can be directly provided to the user to reduce multiple rounds of follow-up questions, so as to enable the user to understand the initial reply content more quickly. When the difficulty level corresponding to the user query content is a second difficulty level, or the user's understanding ability and professional ability are determined to be poor based on the user identity information, the initial reply content can be split, and multiple rounds of follow-up questions can be provided to the user, and the amount of information of the initial reply content provided each time is small, so as to facilitate the user to understand.

[0305] In an optional embodiment, the splitting of the initial reply content can be determined according to the user behavior features and the user identity information contained in the current user interaction data. When the preferred content corresponding to the user behavior features is a first preferred content, and the user's understanding ability and professional ability are determined to be good based on the user identity information, the initial reply content can not be split according to the user behavior features and the user identity information, and detailed and complete reply content can be directly provided to the user to reduce multiple rounds of follow-up questions, so as to enable the user to understand the initial reply content more quickly. When the preferred content corresponding to the user behavior features is a second preferred content, or the user's understanding ability and professional ability are determined to be poor based on the user identity information, the initial reply content can be split according to the user behavior features and the user identity information, multiple rounds of follow-up questions can be provided to the user, and the amount of information of the initial reply content provided each time is small, so as to facilitate the user to understand. In an example, the first preferred content is used to represent content in the same field as the user query content; and the second preferred content is used to represent content in a different field from the user query content.

[0306] In an optional embodiment, the splitting of the initial reply content can be determined according to the vehicle data and the user identity information contained in the current user interaction data. When the operation complexity corresponding to the vehicle data is a first operation complexity, and the user's understanding ability and professional ability are determined to be good based on the user identity information, it is determined that the initial reply content is not split, and detailed and complete reply content can be directly provided to the user to reduce multiple rounds of follow-up questions, so as to enable the user to understand the initial reply content more quickly. When the operation complexity corresponding to the vehicle data is a second operation complexity, or the user's understanding ability and professional ability are determined to be poor based on the user identity information, it is determined that the initial reply content is split, multiple rounds of follow-up questions can be provided to the user, and the amount of information of the initial reply content provided each time is small, so as to facilitate the user to understand.

[0307] Generally, the more vehicle data needed by the user question content, the more complex the process of obtaining the initial reply content needs to be processed, and the more reply content can be provided through multiple rounds of dialogue, that is, the initial reply content is split; the less vehicle data needed by the user question content, the simpler the process of obtaining the initial reply content needs to be processed, and more content does not need to be provided through multiple rounds of dialogue, that is, the initial reply content is not split. Generally, the operation is complex, and the initial reply content is split; the operation is simple, and the initial reply content is not split.

[0308] In an optional embodiment, in the case of splitting the initial reply content, the initial reply content includes at least two first reply contents. In an optional embodiment, splitting the initial reply content can split the initial reply content into at least two first reply contents with a total logical relationship, or can split the initial reply content into at least two first reply contents with a progressive relationship or an inference relationship. The information amount of the initial reply content is greater than the information amount of the first reply content.

[0309] For example, the initial reply content: Dinosaurs disappeared because a long time ago, a big event happened on Earth. Imagine a huge stone flying from the sky, crashing onto the earth with a loud noise, which is a large meteorite. The large stone caused a lot of dust and smoke to block the sky, and the sun's light could not reach the ground, and the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0310] In an optional embodiment, the split first reply content is: the disappearance of dinosaurs is because a small planet hits the earth, causing a dramatic change in the environment, leading to the inability to survive. A first reply content is: a huge stone flies from the sky, crashes onto the earth with a loud noise, which is a large meteorite. The large stone caused a lot of dust and smoke to block the sky, making the earth very cold, leading to the disappearance of dinosaurs. A first reply content is: the earth became very cold, and many plants died. Dinosaurs had no food to eat, so they also slowly disappeared.

[0311] In an optional embodiment, "generating reply information according to the user interaction data and the reply information structure" is refined as: determining a target information amount corresponding to the user interaction data according to the user interaction data; generating reply information according to the user interaction data, the target information amount, and the reply information structure; wherein the reply information at least includes a first reply content; the information amount of the first reply content corresponds to the target information amount.

[0312] In an optional embodiment, FIG. 8 is a flowchart of another method for generating reply information according to an embodiment of the present application. This embodiment is a detailed description of generating reply information according to the user interaction data and the reply information structure on the basis of the above-mentioned embodiments.

[0313] It should be noted that the parts not described in detail in the embodiments of the present application can refer to the descriptions of the above-mentioned embodiments.

[0314] Referring to the method for generating reply information shown in FIG. 8, the method comprises the following steps.

[0315] S801, obtaining user interaction data.

[0316] S802, determining a reply information structure according to the user interaction data.

[0317] S803, determining a target information amount corresponding to the user interaction data according to the user interaction data.

[0318] In an optional embodiment, the target information amount is used to indicate the information amount of the generated first reply content. The target information amount can be a specific numerical value, for example, the number of words. Or the target information amount can be a level or a type. For example, the target information amount is more or less information amount.

[0319] In an optional embodiment, the target information amount can be determined in the following manner: for example, a mapping relationship between the user interaction data and the target information amount can be preset, and the target information amount corresponding to the user interaction data can be queried according to the mapping relationship. For another example, the user interaction data can be input into a pre-trained deep learning model to obtain the target information amount. Specifically, the user interaction data is subjected to feature extraction to obtain a feature vector, and the feature vector is subjected to decoding classification to obtain the target information amount. For another example, the user interaction data is subjected to quantization to obtain the target information amount.

[0320] S804, generating reply information according to the user interaction data, the target information amount and the reply information structure; wherein the reply information at least comprises first reply content; and the information amount of the first reply content corresponds to the target information amount.

[0321] In an optional embodiment, the generation of the reply information according to the user interaction data, the target information amount and the reply information structure can be: inputting the user interaction data, the target information amount and the reply information structure into a pre-trained deep learning model to obtain the reply information output by the deep learning model. In an example, the user interaction data, the target information amount and the reply information structure can be filled into a prompt template corresponding to the target information amount to obtain input data, and the input data can be input into a large language model to obtain the reply information output by the large language model.

[0322] For example, according to the user interaction data, the target information amount and the reply information structure, the generating of the reply information can be: according to the user interaction data and the target information amount, generating reply indication information, and according to the reply indication information, the user interaction data and the reply information structure, generating the reply information. Specifically, input the user interaction data into the first large language model to obtain the target information amount output by the first large language model, and query or generate the reply indication information according to the target information amount. According to the user interaction data, the reply indication information and the reply information structure, determine the input data. Input the input data into the second large language model to obtain the reply information output by the second large language model. The first large language model and the second large language model can be the same or different. Wherein, the first large language model can also be replaced by a classification model.

[0323] The embodiment of the application can obtain the target information amount by processing the user interaction data, and generate the reply information including the first reply content corresponding to the target information amount according to the target information amount, the user interaction data and the reply information structure. The target information amount can be directly extracted from the user interaction data, and the reply information is generated based on the target information amount, the reply information structure and the user interaction data to determine whether the reply information contains the second reply content on the basis of containing the first reply content. The information amount of the reply information can be accurately controlled.

[0324] In an optional embodiment, the determining of the target information amount corresponding to the user interaction data according to the user interaction data comprises: determining the target information amount corresponding to the user interaction data according to the current user interaction data and / or the historical user interaction data. Wherein, the current user interaction data and the historical user interaction data can each include at least one of the following: user attribute information; user question content; user historical dialogue; user behavior characteristics; vehicle data. Wherein, the current user interaction data refers to the data collected in the current user interaction process; the historical user interaction data refers to the data collected in the historical user interaction process.

[0325] In an optional embodiment, the current user attribute information includes a first age; the historical user attribute information includes a second age; the first age and the second age are different, and the first information amount corresponding to the first age and the second information amount corresponding to the second age are different.

[0326] In an optional embodiment, FIG. 9 is a flowchart of another reply information generation method provided by the embodiment of the application.

[0327] In an optional embodiment, the reply information generation method further comprises: dynamically adjusting the associated guidance frequency according to the user interaction data.

[0328] It should be noted that the parts not described in detail in the embodiments of the present application can refer to the descriptions of the foregoing embodiments.

[0329] Referring to the reply information generation method shown in FIG. 9, the method comprises the following steps:

[0330] S901, acquiring user interaction data.

[0331] S902, determining a guidance type corresponding to the user interaction data according to the user interaction data.

[0332] S903, dynamically adjusting an associated guidance frequency according to the user interaction data.

[0333] S904, generating reply information according to the guidance frequency, the user interaction data, and the guidance type; the reply information at least comprises first reply content; and whether the reply information comprises second reply content corresponds to the guidance type.

[0334] The guidance frequency is used to represent the frequency of guiding the user interaction data. In an example, the guidance frequency can be represented by the ratio of the guidance round to the total number of interactions, that is, in the case of a certain total number of interactions, the guidance round is proportional to the guidance frequency. For example, assuming that in the process of user interaction, the total number of interactions is 10 times, and the guidance round is 4 times, then the guidance frequency is 0.4. In an embodiment, the understanding ability and professional ability of the user can be determined based on the user interaction data, so as to dynamically increase or decrease the associated guidance frequency based on the understanding ability and professional ability of the user, thereby improving the user's interaction experience.

[0335] In an optional embodiment, the user interaction data comprises current user interaction data and historical user interaction data; and the dynamically adjusting the associated guidance frequency according to the user interaction data comprises:

[0336] dynamically adjusting the associated guidance frequency according to the current user interaction data and the historical user interaction data.

[0337] The current user interaction data and the historical user interaction data can each comprise at least one of the following: user attribute information; user question content; user historical dialogue; user behavior characteristics; and vehicle data. The current user interaction data refers to the data collected in the current user interaction process, and the historical user interaction data refers to the data collected in the historical user interaction process.

[0338] In an optional embodiment, the current user attribute information comprises a first age, the historical user attribute information comprises a second age, the first age is different from the second age, and a first guidance frequency corresponding to the first age is different from a second guidance frequency corresponding to the second age. In an example, content with a large guidance frequency can be fed back to a user with a small age, so as to reduce the difficulty of understanding the reply content. The first age is less than the second age, and accordingly, the first guidance frequency is less than the second guidance frequency.

[0339] Optionally, when the difficulty level of the current user question content is a first difficulty level, a third guidance frequency corresponding to the current user interaction data is determined according to the current user question content; when the difficulty level of the historical user question content is a second difficulty level, a fourth guidance frequency corresponding to the historical user interaction data is determined according to the historical user question content; and when the first difficulty level is different from the second difficulty level, the third guidance frequency is different from the fourth guidance frequency.

[0340] In an optional embodiment, the difficulty level of the user question content is used to determine the complexity of the answer, so as to determine the guidance frequency. The difficulty level of the user question content can be classified, for example, in terms of difficulty or simplicity. For example, a mapping relationship between the difficulty level and the guidance frequency can be preset, and the corresponding guidance frequency can be determined according to the difficulty level. For another example, the user question content can be input into a deep learning model to obtain the guidance frequency.

[0341] Optionally, when the preferred content corresponding to the current user behavior feature is a first preferred content, a seventh guidance frequency corresponding to the current user interaction data is determined according to the current user behavior feature; when the preferred content corresponding to the historical user behavior feature is a second preferred content, an eighth guidance frequency corresponding to the historical user interaction data is determined according to the historical user behavior feature; and when the first preferred content is different from the second preferred content, the seventh guidance frequency is different from the eighth guidance frequency.

[0342] In an optional embodiment, the user behavior feature is used to determine the preferred content. The guidance frequency is determined according to the preferred content. The guidance frequency is different for different preferred contents. In practice, a user will want more content of interest. Therefore, the guidance frequency can be determined according to the preferred content.

[0343] Optionally, when the operation complexity corresponding to the current vehicle data is a first operation complexity, a ninth guidance frequency corresponding to the current user interaction data is determined according to the current vehicle data; when the operation complexity corresponding to the historical vehicle data is a second operation complexity, a tenth guidance frequency corresponding to the historical user interaction data is determined according to the historical vehicle data; and when the first operation complexity is different from the second operation complexity, the ninth guidance frequency is different from the tenth guidance frequency.

[0344] The operation complexity is used to describe the complexity of processing the first reply content. The user query content and the vehicle data can be questioned, the vehicle data related to the user query content is determined, the related vehicle data is classified, and the operation complexity is determined.

[0345] For different operation complexities, the determined guidance frequency is different. Generally, the more vehicle data needed by the user query content, the more complex the process of processing the first reply content, and the more the first reply content provided accordingly, that is, the guidance frequency is large; the less vehicle data needed by the user query content, the simpler the process of processing the first reply content, and the less the first reply content provided accordingly, that is, the guidance frequency is small. Different user query contents can correspond to different operation complexities. Generally, the guidance frequency is large for high operation complexity, and the guidance frequency is small for low operation complexity. For example, a mapping relationship between the operation complexity and the guidance frequency can be preset, and the corresponding guidance frequency is determined according to the operation complexity. For another example, the operation complexity can be input into a deep learning model to obtain the guidance frequency.

[0346] In an optional embodiment, the dynamically adjusting the associated guidance frequency according to the user interaction data comprises: when the user attribute information is first user interaction data, dynamically increasing the associated guidance frequency according to the first user interaction data; when the user attribute information is second user interaction data, dynamically decreasing the associated guidance frequency according to the second user interaction data.

[0347] In an example, if the first user interaction data is the first user attribute information, indicating that the user's understanding ability and professional ability are poor, the associated guidance frequency can be dynamically increased; if the user interaction data is the second user attribute information, indicating that the user's understanding ability and professional ability are good, the associated guidance frequency can be dynamically decreased, thereby achieving dynamic adjustment of the guidance frequency.

[0348] In an optional embodiment, the first user attribute information comprises a first age, and the second user attribute information comprises a second age. The first age and the second age are different, and the guidance frequency corresponding to the first age and the guidance frequency corresponding to the second age are different. In an example, the second age is greater than the first age. For example, the first age and the second age can be represented by an age range, such as the first age range being below 18 years old, and the second age range being above 18 years old. In an example, when the user's age is within the range of the first age, the corresponding guidance frequency can be increased; when the user's age is within the range of the second age, the corresponding guidance frequency can be decreased, thereby achieving the effect of dynamically adjusting the guidance frequency based on the user attribute information, and thereby the dynamic adjustment of the amount of information contained in the reply information.

[0349] In an optional embodiment, the reply information generation method further comprises: in response to the received guidance stop operation, dynamically switching the associated guidance type. The guidance type comprises a guided type and an unguided type. Whether the reply information comprises the second reply content corresponds to the guidance type, which means that when the guidance type is the guided type, the reply information comprises the second reply content; and when the guidance type is the unguided type, the reply information does not comprise the second reply content. In actual operation, whether to dynamically switch between the guided type and the unguided type can be determined based on the user's interactive operation. In an example, if the guidance type of the user interaction data is the guided type, but during the interaction, the guidance stop operation is received, the guidance type is switched from the guided type to the unguided type, so as to dynamically adjust the amount of information contained in the reply information.

[0350] In an optional embodiment, the guidance stop operation comprises at least one of the following: a guidance rejection operation; no guidance reply content is received within a preset valid time length; information unrelated to the user question content is received. In an example, a guidance type switch icon can be configured and displayed on the user interaction interface, and the configuration options of the guidance type can be intuitively displayed to the user on the user interaction interface. If the user can directly turn off the guidance type switch icon on the user interaction interface, that is, the user interaction interface receives the guidance rejection operation, at this time the guidance type is set to the unguided type, and the second reply content is not contained in the reply information. In an example, if the guidance reply content provided by the user is not received within the preset valid time length when generating the reply information, at this time it can be considered that there is no need for guidance, then the guidance type is switched from the guided type to the unguided type, and only the first reply content is contained in the reply information generated next time, thereby reducing the amount of information and calculation of the reply information. The preset valid time length is used to represent the waiting time length of the guidance content after receiving the reply information, for example, the preset valid time length is 3s, that is, after generating and outputting the reply information to the user, if the user's reply guidance content is not received within 3s, at this time the guidance type is switched from the guided type to the unguided type. In an example, the information unrelated to the user question content refers to the information that the user's reply content is unrelated to the user question content. For example, assuming that the user question content is: "Why is the earth round?", the reply is: "The earth's gravity is an important reason for the formation of the earth's sphere. Do you want to know the specific implementation process of the formation of the earth's sphere?", if the user replies "Why is the moon sometimes curved?", at this time, it can be determined that the user generally will not ask follow-up questions, and the guidance type can be switched from the guided type to the unguided type to improve the user's question experience.

[0351] In an optional embodiment, the reply information generation method further comprises: determining the associated actual guiding round and the guiding sentence pattern associated with each actual guiding round according to the user interaction data and the user attribute information. In an example, the actual guiding round is used to represent the total number of guiding types in one interaction process; the guiding sentence pattern is used to represent the format associated with the guiding content, and the guiding sentence pattern can also be used to guide whether the user wants to terminate the guiding process, for example, the guiding sentence pattern can include: guiding the user to continue to ask questions; guiding the user to terminate the question.

[0352] In an example, the user interaction data can include: user question content, and the actual guiding round and the guiding sentence pattern associated with each actual guiding round can be determined based on the difficulty level corresponding to the user question content and the user attribute information. If the difficulty level corresponding to the user question content is a first difficulty level, and it is determined based on the user attribute information that the user's understanding ability and professional ability are poor, then the associated actual guiding round is determined to be larger, and the guiding sentence pattern associated with each actual guiding round is a sentence pattern guiding the user to continue to ask questions; if the difficulty level corresponding to the user question content is a second difficulty level, or it is determined based on the user attribute information that the user's understanding ability and professional ability are good, then the associated actual guiding round is determined to be smaller, and the guiding sentence pattern associated with each actual guiding round is a sentence pattern guiding the user to terminate the question, so as to directly provide detailed and complete answers to the user, and reduce the cumbersome process of repeated questions by the user.

[0353] In an example, the user interaction data can include: user historical dialogue, and the actual guiding round and the guiding sentence pattern associated with each actual guiding round can be determined based on the user historical dialogue and the user attribute information. If the user historical dialogue is first known information, and it is determined based on the user attribute information that the user's understanding ability and professional ability are poor, then the associated actual guiding round is determined to be larger, and the guiding sentence pattern associated with each actual guiding round is a sentence pattern guiding the user to continue to ask questions; if the user historical dialogue is first known information, or it is determined based on the user attribute information that the user's understanding ability and professional ability are good, then the associated actual guiding round is determined to be smaller, and the guiding sentence pattern associated with each actual guiding round is a sentence pattern guiding the user to terminate the question, so as to directly provide detailed and complete answers to the user, and reduce the cumbersome process of repeated questions by the user. In an example, the first known information and the second known information are different.

[0354] In an example, the user interaction data can include user behavior features, and the actual guiding round number and the guiding sentence pattern associated with each actual guiding round number can be determined based on the user behavior features and the user attribute information. If the preference content corresponding to the user behavior features is first preference content, and the understanding ability and the professional ability of the user are determined to be poor based on the user attribute information, the actual guiding round number associated is determined to be larger, and the guiding sentence pattern associated with each actual guiding round number is a sentence pattern guiding the user to continue to ask questions; if the preference content corresponding to the user behavior features is second preference content, or the understanding ability and the professional ability of the user are determined to be good based on the user attribute information, the actual guiding round number associated is determined to be smaller, and the guiding sentence pattern associated with each actual guiding round number is a sentence pattern guiding the user to terminate the question, so as to directly provide detailed and complete answers to the user, and reduce the tedious process of repeated questions of the user. In an example, the first preference content is different from the second preference content.

[0355] In an example, the user interaction data can include vehicle data, and the actual guiding round number and the guiding sentence pattern associated with each actual guiding round number can be determined based on the operation complexity corresponding to the vehicle data and the user attribute information. If the operation complexity corresponding to the vehicle data is first operation complexity, and the understanding ability and the professional ability of the user are determined to be poor based on the user attribute information, the actual guiding round number associated is determined to be larger, and the guiding sentence pattern associated with each actual guiding round number is a sentence pattern guiding the user to continue to ask questions; if the operation complexity corresponding to the vehicle data is second operation complexity, or the understanding ability and the professional ability of the user are determined to be good based on the user attribute information, the actual guiding round number associated is determined to be smaller, and the guiding sentence pattern associated with each actual guiding round number is a sentence pattern guiding the user to terminate the question, so as to directly provide detailed and complete answers to the user, and reduce the tedious process of repeated questions of the user. In an example, the first operation complexity is different from the second operation complexity.

[0356] In an optional embodiment, the user interaction data includes current user interaction data and historical user interaction data; and the actual guiding round number and the guiding sentence pattern associated with each actual guiding round number are determined based on the current user interaction data and the historical user interaction data.

[0357] If the user question content contained in the current user interaction data is based on certain knowledge of the user, at this time, and the user attribute information is the age of the user and is within the first age range, at this time, without multiple guidance to the user, the actual guidance round is set to a value less than the preset guidance round threshold, and the guidance sentence is set to the first guidance sentence, and the user is guided to terminate the question, thereby facilitating the direct provision of detailed and complete answers to the user, and reducing the tedious process of repeated questioning by the user.

[0358] In an optional embodiment, the determination of the associated actual guidance round and the guidance sentence associated with each actual guidance round according to the user interaction data comprises: when the user interaction data is the first user interaction data, determining that the associated actual guidance round is less than the preset guidance round threshold, and the guidance sentence associated with each actual guidance round is the first guidance sentence; when the user interaction data is the second user interaction data, determining that the associated actual guidance round is greater than the preset guidance round threshold, and the guidance sentence associated with each actual guidance round is the second guidance sentence. In an optional embodiment, when the user question content contained in the user interaction data matches the third user attribute information, and the user attribute information includes the second user attribute information, it is determined that the associated actual guidance round is less than the preset guidance round threshold, and the guidance sentence associated with each actual guidance round is the first guidance sentence; when the user question content contained in the user interaction data does not match the third user attribute information, and the user attribute information includes the first user attribute information, it is determined that the associated actual guidance round is greater than the preset guidance round threshold, and the guidance sentence associated with each actual guidance round is the second guidance sentence.

[0359] In an example, the first guidance sentence is different from the second guidance sentence, for example, the first guidance sentence can be to guide the user to terminate the question; and the second guidance sentence can be to guide the user to continue the question.

[0360] In an example, the third attribute information is used to represent relevant information matching the user's own occupation and field of interest. In an example, if the user question content contained in the user interaction data matches the third user attribute information, it indicates that the user has certain knowledge of the user question content, at this time, if the user attribute information includes the second user attribute information, i.e., the age of the user is within the second age range, at this time, without multiple guidance to the user, the actual guidance round is set to a value less than the preset guidance round threshold, and the guidance sentence is set to the first guidance sentence, and the user is guided to terminate the question, thereby facilitating the direct provision of detailed and complete answers to the user, and reducing the tedious process of repeated questioning by the user.

[0361] In an example, if the user interaction data contains user question content that does not match the third user attribute information, it is indicated that the user has little understanding of the user question content, or if the user attribute information includes the first user attribute information, i.e., the user's age is within the first age range, i.e., the user's understanding ability and professional ability are relatively poor, at this time, the user needs to be guided multiple times, the actual guidance round is set to a value greater than the preset guidance round threshold, and the guidance sentence pattern is set to the second guidance sentence pattern, and the user is guided to continue to ask questions to satisfy the user's curiosity.

[0362] FIG. 10 is a flowchart of another reply information generation method according to an embodiment of the present application. The embodiment can be applied to a case of replying to a user's question in a human-computer interaction process between a vehicle and a user. The method can be executed by a reply information generation device.

[0363] Referring to the reply information generation method shown in FIG. 10, the method includes the following steps.

[0364] S1001, obtaining user interaction data.

[0365] S1002, determining a reply information structure according to the user interaction data.

[0366] S1003, generating reply information according to the user interaction data and the reply information structure.

[0367] It should be noted that the parts not described in detail in the embodiments of the present application can be referred to the descriptions of the foregoing embodiments.

[0368] The embodiments of the present application can flexibly adjust the reply information structure of the reply information by determining the reply information structure matched with the user interaction data and generating the reply information corresponding to the user interaction data according to the reply information structure, thereby effectively improving the human-computer interaction experience.

[0369] FIG. 11 is a flowchart of another reply information generation method according to an embodiment of the present application. The embodiment can be applied to a case of replying to a user's question in a human-computer interaction process between a vehicle and a user. The method can be executed by a reply information generation device.

[0370] Referring to the reply information generation method shown in FIG. 11, the method includes the following steps.

[0371] S1101, obtaining user interaction data.

[0372] S1102, determining a guidance type corresponding to the user interaction data according to the user interaction data.

[0373] S1103, generating reply information according to the user interaction data and the guidance type; the reply information at least includes first reply content; whether the reply information includes the second reply content corresponds to the guidance type.

[0374] It should be noted that the parts not described in detail in the embodiments of the present application can refer to the descriptions of the foregoing embodiments.

[0375] The embodiments of the present application obtain the guidance type by processing the user interaction data, and generate the reply information including or not including the second reply content according to the guidance type and the user interaction data, which can directly extract the guidance type from the user interaction data and generate the reply information based on the guidance type and the user interaction data, so as to accurately control whether to generate the second reply content and flexibly guide the user to ask questions, thereby effectively improving the human-computer interaction experience.

[0376] FIG. 12 is a structural schematic diagram of a reply information generation device provided by an embodiment of the present application. The embodiments of the present application can be applied to the case of replying to the user's question in the human-computer interaction process between the vehicle and the user. The device can execute the reply information generation method. The reply information generation device can be realized in the form of hardware and / or software. The device can be configured in an electronic device.

[0377] Referring to the reply information generation device shown in FIG. 12, it includes:

[0378] An interaction data acquisition module 1201 is configured to acquire user interaction data.

[0379] A reply information generation module 1202 is configured to generate reply information according to the user interaction data, wherein the reply information at least includes first reply content, and the information amount of the first reply content corresponds to the user interaction data.

[0380] The embodiments of the present application generate the reply information with an information amount adapted to the user interaction data, which can flexibly adjust the information amount of the reply information and specifically realize the increase or decrease of the information amount of the reply information, so as to feed back the reply information convenient for the user to understand.

[0381] Optionally, the reply information generation module 1202 includes:

[0382] A target information amount determination unit is configured to determine a target information amount corresponding to the user interaction data according to the user interaction data.

[0383] A first reply content generation unit is configured to generate reply information according to the user interaction data and the target information amount, wherein the information amount of the first reply content corresponds to the target information amount.

[0384] Optionally, the user interaction data comprises user attribute information; the target information amount determination unit is specifically configured to: when the user attribute information is first user attribute information, determine a first information amount corresponding to the first user attribute information according to the first user attribute information; when the user attribute information is second user attribute information, determine a second information amount corresponding to the second user attribute information according to the second user attribute information; the first user attribute information is different from the second user attribute information, and the first information amount is different from the second information amount.

[0385] Optionally, the first user attribute information comprises a first age; the second user attribute information comprises a second age; the first age is different from the second age, and a first information amount corresponding to the first age is different from a second information amount corresponding to the second age.

[0386] Optionally, the user interaction data comprises user question content; the target information amount determination unit is specifically configured to: when a difficulty level corresponding to the user question content is a first difficulty level, determine a third information amount corresponding to the user interaction data according to the user question content; when a difficulty level corresponding to the user question content is a second difficulty level, determine a fourth information amount corresponding to the user interaction data according to the user question content; when the first difficulty level is different from the second difficulty level, the third information amount is different from the fourth information amount.

[0387] Optionally, the user interaction data comprises user historical dialogue; the target information amount determination unit is specifically configured to: when known information corresponding to the user historical dialogue is first known information, determine a fifth information amount corresponding to the user interaction data according to the user historical dialogue; when known information corresponding to the user historical dialogue is second known information, determine a sixth information amount corresponding to the user interaction data according to the user historical dialogue; when the first known information is different from the second known information, the fifth information amount is different from the sixth information amount.

[0388] Optionally, the user interaction data comprises user behavior features; the target information amount determination unit is specifically configured to: when preference content corresponding to the user behavior features is first preference content, determine a seventh information amount corresponding to the user interaction data according to the user behavior features; when preference content corresponding to the user behavior features is second preference content, determine an eighth information amount corresponding to the user interaction data according to the user behavior features; when the first preference content is different from the second preference content, the seventh information amount is different from the eighth information amount.

[0389] Optionally, the user interaction data comprises vehicle data; the target information amount determination unit is specifically configured to: when an operation complexity corresponding to the vehicle data is a first operation complexity, determine a ninth information amount corresponding to the user interaction data according to the vehicle data; when the operation complexity corresponding to the vehicle data is a second operation complexity, determine a tenth information amount corresponding to the user interaction data according to the vehicle data; and when the first operation complexity is different from the second operation complexity, the ninth information amount is different from the tenth information amount.

[0390] Optionally, the reply information further comprises second reply content, and the second reply content is guided follow-up question content corresponding to the user interaction data.

[0391] Optionally, the reply information generation module 1202 comprises:

[0392] a guide type determination unit configured to determine a guide type corresponding to the user interaction data according to the user interaction data;

[0393] a second reply content determination unit configured to generate reply information according to the user interaction data and the guide type, and whether the reply information comprises second reply content corresponding to the guide type.

[0394] Optionally, the user interaction data comprises user attribute information; the guide type determination unit is specifically configured to: when the user attribute information is first user attribute information, determine a first guide type corresponding to the first user attribute information according to the first user attribute information; when the user attribute information is second user attribute information, determine a second guide type corresponding to the second user attribute information according to the second user attribute information; and the first user attribute information is different from the second user attribute information, and the first guide type is different from the second guide type.

[0395] Optionally, the user interaction data comprises user question content; the guide type determination unit is specifically configured to: when a difficulty level corresponding to the user question content is a first difficulty level, determine a third guide type corresponding to the user interaction data according to the user question content; when the difficulty level corresponding to the user question content is a second difficulty level, determine a fourth guide type corresponding to the user interaction data according to the user question content; and when the first difficulty level is different from the second difficulty level, the third guide type is different from the fourth guide type.

[0396] Optionally, the user interaction data comprises: user historical conversation; the guidance type determination unit is specifically configured to: when the known information corresponding to the user historical conversation is first known information, determine a fifth guidance type corresponding to the user interaction data according to the user historical conversation; when the known information corresponding to the user historical conversation is second known information, determine a sixth guidance type corresponding to the user interaction data according to the user historical conversation; and when the first known information is different from the second known information, the fifth guidance type is different from the sixth guidance type.

[0397] Optionally, the user interaction data comprises: user behavior characteristics; the guidance type determination unit is specifically configured to: when the preference content corresponding to the user behavior characteristics is first preference content, determine a seventh guidance type corresponding to the user interaction data according to the user behavior characteristics; when the preference content corresponding to the user behavior characteristics is second preference content, determine an eighth guidance type corresponding to the user interaction data according to the user behavior characteristics; and when the first preference content is different from the second preference content, the seventh guidance type is different from the eighth guidance type.

[0398] Optionally, the user interaction data comprises: vehicle data; the guidance type determination unit is specifically configured to: when the operation complexity corresponding to the vehicle data is first operation complexity, determine a ninth guidance type corresponding to the user interaction data according to the vehicle data; when the operation complexity corresponding to the vehicle data is second operation complexity, determine a tenth guidance type corresponding to the user interaction data according to the vehicle data; and when the first operation complexity is different from the second operation complexity, the ninth guidance type is different from the tenth guidance type.

[0399] The reply information generation device provided by the embodiment of the application can execute the reply information generation method provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of executing the reply information generation method.

[0400] FIG. 13 is a schematic diagram of another kind of reply information generation device provided by the embodiment of the application. The embodiment of the application can be applied to the case of replying to the user's question in the human-computer interaction process between the vehicle and the user. The device can execute the reply information generation method, and the reply information generation device can be realized in the form of hardware and / or software, and the device can be configured in an electronic device.

[0401] Referring to the reply information generation device shown in FIG. 13, the device comprises:

[0402] An interaction data acquisition module 1301 is configured to acquire user interaction data.

[0403] The reply information structure determination module 1302 is configured to determine a reply information structure according to the user interaction data.

[0404] The reply information generation module 1303 is configured to generate reply information according to the user interaction data and the reply information structure.

[0405] The embodiment of the present application can flexibly adjust the reply information structure of the reply information by determining the reply information structure matched with the user interaction data and generating the reply information corresponding to the user interaction data according to the reply information structure, thereby effectively improving the human-computer interaction experience.

[0406] Optionally, the reply information structure comprises: a containing condition of the first reply content and / or the second reply content.

[0407] Optionally, the first reply content corresponds to the user interaction data.

[0408] Optionally, the second reply content is a guided follow-up question content corresponding to the user interaction data.

[0409] Optionally, the reply information structure determination module 1302 is specifically configured to determine the reply information structure according to current user interaction data and / or historical user interaction data.

[0410] The reply information structure determination module 1302 is specifically configured to, when the user interaction data is first user interaction data, determine, according to the first user interaction data, that a reply information structure corresponding to the first user interaction data is a first reply information structure.

[0411] When the user interaction data is second user interaction data, the reply information structure determination module 1302 is specifically configured to determine, according to the second user interaction data, that a reply information structure corresponding to the second user interaction data is a second reply information structure.

[0412] Optionally, the reply information generation apparatus further comprises:

[0413] The initial reply content determination module is configured to determine an initial reply content according to the user interaction data.

[0414] The split condition determination module is configured to determine a split condition of the initial reply content according to the user interaction data.

[0415] Optionally, the split condition determination module is specifically configured to:

[0416] determine the split condition of the initial reply content according to current user interaction data and / or historical user interaction data.

[0417] Optionally, the splitting condition determining module is specifically configured to: when the user interaction data is first user interaction data, determine, according to the first user interaction data, that the splitting condition of the initial reply content is splitting; and when the user interaction data is second user interaction data, determine, according to the second user interaction data, that the splitting condition of the initial reply content is not splitting.

[0418] Optionally, when the splitting condition of the initial reply content is splitting, the initial reply content comprises at least two first reply contents.

[0419] Optionally, the reply information generation module 1303 comprises:

[0420] A target information amount determining unit is configured to determine a target information amount corresponding to the user interaction data according to the user interaction data.

[0421] A first reply content generation unit is configured to generate reply information according to the user interaction data, the target information amount and the reply information structure, wherein the reply information at least comprises a first reply content, and an information amount of the first reply content corresponds to the target information amount.

[0422] Optionally, the target information amount determining unit is specifically configured to determine the target information amount corresponding to the user interaction data according to current user interaction data and / or historical user interaction data.

[0423] Optionally,

[0424] The target information amount determining unit is specifically configured to: when the user attribute information is first user interaction data, determine, according to the first user interaction data, that a target information amount corresponding to the first user interaction data is a first information amount; and when the user interaction data is second user interaction data, determine, according to the second user interaction data, that a target information amount corresponding to the second user interaction data is a second information amount.

[0425] Optionally, the user interaction data comprises at least one of the following: user attribute information; user question content; user historical dialogue; user behavior characteristics; and vehicle data.

[0426] The reply information generation device provided by the embodiment of the application can execute the reply information generation method provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of executing the reply information generation method.

[0427] FIG. 14 is a structural schematic diagram of another reply information generation device provided by an embodiment of the present application. The embodiment of the present application can be applied to the case of replying to the user's question in the process of human-computer interaction between the vehicle and the user. The device can execute the reply information generation method. The reply information generation device can be realized in the form of hardware and / or software. The device can be configured in an electronic device.

[0428] Referring to the reply information generation device shown in FIG. 14, the device comprises:

[0429] An interaction data acquisition module 1401 is configured to acquire user interaction data.

[0430] A guide type determination module 1402 is configured to determine a guide type corresponding to the user interaction data according to the user interaction data.

[0431] A reply information generation module 1403 is configured to generate reply information according to the user interaction data and the guide type. The reply information at least comprises first reply content. Whether the reply information comprises second reply content corresponds to the guide type.

[0432] The embodiment of the present application generates reply information with an information amount adapted to the user interaction data. The information amount of the reply information can be flexibly adjusted. The information amount of the reply information can be specifically increased or decreased. Thus, the reply information is easy for the user to understand.

[0433] Optionally, the information amount of the first reply content corresponds to the user interaction data. The second reply content is guide follow-up question content corresponding to the user interaction data.

[0434] Optionally, the guide type determination module 1402 is specifically configured to: when the user interaction data is first user interaction data, determine a first information amount corresponding to the first user interaction data according to the first user interaction data; and when the user interaction data is second user interaction data, determine a second information amount corresponding to the second user interaction data according to the second user interaction data.

[0435] Optionally, the user interaction data comprises user attribute information.

[0436] The guide type determination module 1402 is specifically configured to: when the user interaction data is first user interaction data, determine a first guide type corresponding to the first user interaction data according to the first user interaction data; and when the user interaction data is second user interaction data, determine a second guide type corresponding to the second user interaction data according to the second user interaction data.

[0437] In an optional embodiment, the reply information generation device further comprises:

[0438] The guide frequency adjustment module is configured to dynamically adjust the associated guide frequency according to the user attribute information.

[0439] In an optional embodiment, the user interaction data comprises current user interaction data and historical user interaction data; and the guide frequency adjustment module is specifically configured to:

[0440] dynamically adjust the associated guide frequency according to the current user interaction data and the historical user interaction data.

[0441] In an optional embodiment, the user interaction data comprises first user interaction data and second user interaction data; and the guide frequency adjustment module is specifically configured to: when the user interaction data is the first user interaction data, dynamically increase the associated guide frequency according to the first user interaction data; and when the user interaction data is the second user interaction data, dynamically decrease the associated guide frequency according to the second user interaction data.

[0442] In an optional embodiment, the reply information generation apparatus further comprises:

[0443] The guide type switching module is configured to dynamically switch the associated guide type in response to the received guide stop operation.

[0444] In an optional embodiment, the guide stop operation comprises at least one of the following: a guide rejection operation; no guide reply content is received within a preset valid time length; information irrelevant to the user question content is received.

[0445] In an optional embodiment, the reply information generation apparatus further comprises a guide round and sentence determination module configured to determine the actual guide round associated and the guide sentence associated with each actual guide round according to the user interaction data and the user attribute information.

[0446] In an optional embodiment, the user interaction data comprises current user interaction data and historical user interaction data; and the guide round and sentence determination module is specifically configured to determine the actual guide round associated and the guide sentence associated with each actual guide round according to the current user interaction data and the historical user interaction data.

[0447] In an optional embodiment, the guiding round and sentence determining module is specifically configured to: when the user interaction data is first user interaction data, determine that the associated actual guiding round is less than a preset guiding round threshold, and the guiding sentence associated with each actual guiding round is a first guiding sentence; when the user interaction data is second user interaction data, determine that the associated actual guiding round is greater than the preset guiding round threshold, and the guiding sentence associated with each actual guiding round is a second guiding sentence.

[0448] Optionally, the reply information generating module 1403 comprises a target information amount determining unit configured to determine a target information amount corresponding to the user interaction data according to the user interaction data.

[0449] The first reply content generating unit is configured to generate reply information according to the user interaction data and the target information amount, and the information amount of the first reply content corresponds to the target information amount.

[0450] Optionally, the user interaction data comprises current user interaction data and historical user interaction data, and the target information amount determining unit is specifically configured to determine the target information amount corresponding to the user interaction data according to the current user interaction data and / or the historical user interaction data.

[0451] Optionally, the user interaction data comprises first user interaction data and second user interaction data, and the target information amount determining unit is specifically configured to: when the user interaction data is the first user interaction data, determine a first information amount corresponding to the first user interaction data according to the first user interaction data; and when the user interaction data is the second user interaction data, determine a second information amount corresponding to the second user interaction data according to the second user interaction data.

[0452] In an optional embodiment, the user interaction data comprises at least one of the following: user attribute information; user question content; user historical conversation; user behavior characteristics; and vehicle data.

[0453] The reply information generating device provided by the embodiments of the present application can execute the reply information generating method provided by any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of executing the reply information generating method.

[0454] FIG. 15 illustrates a structural diagram of an electronic device 1500 that can be used to implement embodiments of the present application. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit the implementations of the present application described and / or claimed in this document.

[0455] As shown in FIG. 15, the electronic device 1500 includes at least one processor 1501, and memory, such as read-only memory (ROM) 1502, random access memory (RAM) 1503, etc., that is communicatively connected to the at least one processor 1501, where the memory stores computer programs that can be executed by the at least one processor. The processor 1501 can perform various appropriate actions and processes according to the computer programs stored in the read-only memory (ROM) 1502 or loaded into the random access memory (RAM) 1503 from the storage unit 1508. In the RAM 1503, various programs and data required for the operation of the electronic device 1500 can also be stored. The processor 1501, the ROM 1502, and the RAM 1503 are connected to each other through a bus 1504. An input / output (I / O) interface 1505 is also connected to the bus 1504.

[0456] Various components in the electronic device 1500 are connected to the I / O interface 1505, including an input unit 1506, such as a keyboard, a mouse, etc., an output unit 1507, such as various types of displays, speakers, etc., a storage unit 1508, such as a magnetic disk, an optical disk, etc., and a communication unit 1509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1509 allows the electronic device 1500 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0457] The processor 1501 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the processor 1501 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 1501 performs various methods and processes described above, such as the reply information generation method.

[0458] In some embodiments, the reply information generation method can be implemented as a computer program tangibly embodied in a computer readable storage medium, e.g., storage unit 1508. In some embodiments, parts or all of the computer program can be loaded and / or installed onto electronic device 1500 via, e.g., ROM 1502 and / or communication unit 1509. When the computer program is loaded onto RAM 1503 and executed by processor 1501, one or more steps of the above-described reply information generation method can be performed. Alternatively, in other embodiments, processor 1501 can be configured to perform the reply information generation method by other any suitable means, e.g., by way of firmware.

[0459] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0460] Computer programs used to implement the methods of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program, when executed by the processor, implements the functions / acts specified in the flowcharts and / or block diagrams. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as part of a standalone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0461] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer-readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of a machine-readable storage medium will include one or more lines of a program of instructions in a transitory signal, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0462] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0463] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0464] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS (Virtual Private Server) service.

[0465] It should be understood that the various forms of flow shown above can be reordered, added to, or have steps deleted. For example, the steps described in the present application can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions of the present application can be achieved, which are not limited herein.

[0466] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A reply information generating method characterized by comprising: The method comprises: obtaining user interaction data; generating reply information according to the user interaction data, the reply information at least comprising first reply content, the information amount of the first reply content corresponding to the user interaction data.

2. The method of claim 1, wherein, The method comprises: determining at least one of a target information amount, a guide type, and a reply information structure corresponding to the user interaction data according to the user interaction data; generating reply information according to at least one of the target information amount, the guide type, and the reply information structure, and the user interaction data.

3. The method of claim 2, wherein, The method comprises: determining at least one of a target information amount, a guide type, and a reply information structure corresponding to the user interaction data according to at least one of user attribute information, a difficulty level corresponding to user question content, known information corresponding to user historical dialogue, preferred content corresponding to user behavior characteristics, and operation complexity corresponding to vehicle data.

4. The method according to claim 2 or 3, characterized in that, The user interaction data comprises user attribute information. The method comprises: when the user attribute information is first user attribute information, determining a first information amount corresponding to the first user attribute information according to the first user attribute information; when the user attribute information is second user attribute information, determining a second information amount corresponding to the second user attribute information according to the second user attribute information; the first user attribute information is different from the second user attribute information, and the first information amount is different from the second information amount.

5. The method of claim 4, wherein, The first user attribute information comprises a first age, and the second user attribute information comprises a second age; the first age is different from the second age, and a first information amount corresponding to the first age is different from a second information amount corresponding to the second age.

6. The method according to any one of claims 2 to 5, characterized in that, The user interaction data comprises user question content. The method comprises: when a difficulty level corresponding to the user question content is a first difficulty level, determining a third information amount corresponding to the user interaction data according to the user question content; when a difficulty level corresponding to the user question content is a second difficulty level, determining a fourth information amount corresponding to the user interaction data according to the user question content; the third information amount is different from the fourth information amount when the first difficulty level is different from the second difficulty level.

7. The method according to any one of claims 2 to 6, characterized in that, The user interaction data comprises user historical dialogue. The method comprises: when known information corresponding to the user historical dialogue is first known information, determining a fifth information amount corresponding to the user interaction data according to the user historical dialogue; when known information corresponding to the user historical dialogue is second known information, determining a sixth information amount corresponding to the user interaction data according to the user historical dialogue; the fifth information amount is different from the sixth information amount when the first known information is different from the second known information. The sixth information amount corresponding to the user interaction data is determined according to the user historical conversation when the known information corresponding to the user historical conversation is second known information; and the fifth information amount is different from the sixth information amount when the first known information is different from the second known information.

8. The method according to any one of claims 2 to 7, characterized in that, The user interaction data comprises user behavior characteristics; The target information amount corresponding to the user interaction data is determined according to the user interaction data, comprising: The seventh information amount corresponding to the user interaction data is determined according to the user behavior characteristics when the preference content corresponding to the user behavior characteristics is first preference content; The eighth information amount corresponding to the user interaction data is determined according to the user behavior characteristics when the preference content corresponding to the user behavior characteristics is second preference content; and the seventh information amount is different from the eighth information amount when the first preference content is different from the second preference content.

9. The method according to any one of claims 2 to 8, characterized in that, The user interaction data comprises vehicle data; The target information amount corresponding to the user interaction data is determined according to the user interaction data, comprising: The ninth information amount corresponding to the user interaction data is determined according to the vehicle data when the operation complexity corresponding to the vehicle data is first operation complexity; The tenth information amount corresponding to the user interaction data is determined according to the vehicle data when the operation complexity corresponding to the vehicle data is second operation complexity; and the ninth information amount is different from the tenth information amount when the first operation complexity is different from the second operation complexity.

10. The method according to any one of claims 2 to 9, characterized in that, Whether the reply information comprises second reply content corresponding to the guidance type, the second reply content being guidance follow-up question content corresponding to the user interaction data.

11. The method according to any one of claims 2 to 10, characterized in that, The user interaction data comprises user attribute information; The guidance type corresponding to the user interaction data is determined according to the user interaction data, comprising: The first guidance type corresponding to the first user attribute information is determined according to the first user attribute information when the user attribute information is first user attribute information; The second guidance type corresponding to the second user attribute information is determined according to the second user attribute information when the user attribute information is second user attribute information; the first user attribute information is different from the second user attribute information, and the first guidance type is different from the second guidance type.

12. The method according to any one of claims 2 to 11, characterized in that, The user interaction data comprises user question content; The guidance type corresponding to the user interaction data is determined according to the user interaction data, comprising: The third guidance type corresponding to the user interaction data is determined according to the user question content when the difficulty level corresponding to the user question content is first difficulty level; The fourth guidance type corresponding to the user interaction data is determined according to the user question content when the difficulty level corresponding to the user question content is second difficulty level; and the third guidance type is different from the fourth guidance type when the first difficulty level is different from the second difficulty level.

13. The method according to any one of claims 2 to 12, characterized in that, The user interaction data comprises user historical conversation; The method further comprises: determining, according to the user interaction data, a guidance type corresponding to the user interaction data, wherein the guidance type comprises at least one of the following: a first guidance type, a second guidance type, a third guidance type, a fourth guidance type, a fifth guidance type, a sixth guidance type, a seventh guidance type, an eighth guidance type, a ninth guidance type, and a tenth guidance type. The method further comprises:

14. The method according to any one of claims 2 to 13, characterized in that, when the known information corresponding to the user historical dialogue is first known information, determining, according to the user historical dialogue, a fifth guidance type corresponding to the user interaction data; when the known information corresponding to the user historical dialogue is second known information, determining, according to the user historical dialogue, a sixth guidance type corresponding to the user interaction data; and when the first known information is different from the second known information, the fifth guidance type is different from the sixth guidance type. The user interaction data comprises: user behavior features. The method further comprises:

15. The method according to any one of claims 2 to 14, characterized in that, when the preferred content corresponding to the user behavior features is first preferred content, determining, according to the user behavior features, a seventh guidance type corresponding to the user interaction data; when the preferred content corresponding to the user behavior features is second preferred content, determining, according to the user behavior features, an eighth guidance type corresponding to the user interaction data; and when the first preferred content is different from the second preferred content, the seventh guidance type is different from the eighth guidance type. The user interaction data comprises: vehicle data. The method further comprises:

16. The method of any one of claims 2-15, wherein, when the operation complexity corresponding to the vehicle data is first operation complexity, determining, according to the vehicle data, a ninth guidance type corresponding to the user interaction data; when the operation complexity corresponding to the vehicle data is second operation complexity, determining, according to the vehicle data, a tenth guidance type corresponding to the user interaction data; and when the first operation complexity is different from the second operation complexity, the ninth guidance type is different from the tenth guidance type.

17. The method of claim 16, wherein, The method further comprises: dynamically adjusting, according to user interaction data, a guidance frequency associated with the guidance type.

18. The method of claim 16, wherein, The user interaction data comprises: current user interaction data and historical user interaction data; and the dynamically adjusting, according to user interaction data, a guidance frequency associated with the guidance type comprises: dynamically adjusting, according to the current user interaction data and the historical user interaction data, the guidance frequency associated with the guidance type. The user interaction data comprises: first user interaction data and second user interaction data; and the dynamically adjusting, according to user interaction data, a guidance frequency associated with the guidance type comprises:

19. The method of any one of claims 2-18, wherein, when the user interaction data is the first user interaction data, dynamically increasing, according to the first user interaction data, the guidance frequency associated with the guidance type; when the user interaction data is the second user interaction data, dynamically decreasing, according to the second user interaction data, the guidance frequency associated with the guidance type.

20. The method of claim 19, wherein, The method further comprises:

21. The method of claims 2-18, wherein, in response to a received guidance stop operation, dynamically switching a guidance type associated with the guidance type. The guidance stop operation comprises at least one of the following: a guidance rejection operation, no guidance reply content being received within a preset valid time length, and information irrelevant to user questioning content being received. The method further comprises: Determine the associated actual guiding round and the guiding sentence pattern associated with each actual guiding round according to the user interaction data.

22. The method of claim 21, wherein, The user interaction data includes current user interaction data and historical user interaction data; and the determining the associated actual guiding round and the guiding sentence pattern associated with each actual guiding round according to the user interaction data includes: Determine the associated actual guiding round and the guiding sentence pattern associated with each actual guiding round according to the current user interaction data and the historical user interaction data.

23. The method of claim 21, wherein, The determining the associated actual guiding round and the guiding sentence pattern associated with each actual guiding round according to the user interaction data includes: When the user interaction data is first user interaction data, determine that the associated actual guiding round is less than a preset guiding round threshold, and the guiding sentence pattern associated with each actual guiding round is a first guiding sentence pattern; When the user interaction data is second user interaction data, determine that the associated actual guiding round is greater than the preset guiding round threshold, and the guiding sentence pattern associated with each actual guiding round is a second guiding sentence pattern.

24. The method of any one of claims 1-23, wherein, The user interaction data includes at least one of the following: user attribute information; user question content; user historical dialogue; user behavior characteristics; and vehicle data.

25. The method of any one of claims 2 to 24, wherein, The reply information structure includes the inclusion of the first reply content and / or the second reply content.

26. The method of any one of claims 2 to 25, wherein, The determining the reply information structure according to the user interaction data includes: Determine the reply information structure and / or the target information amount according to the current user interaction data and / or the historical user interaction data.

27. The method of any one of claims 2 to 25, wherein, The determining the reply information structure according to the user interaction data includes: When the user interaction data is first user interaction data, determine that the reply information structure corresponding to the first user interaction data is a first reply information structure according to the first user interaction data; When the user interaction data is second user interaction data, determine that the reply information structure corresponding to the second user interaction data is a second reply information structure according to the second user interaction data.

28. The method of any one of claims 2 to 25, wherein, The method further includes: Determine initial reply content according to the user interaction data; Determine the splitting condition of the initial reply content according to the user interaction data.

29. The method of claim 28, wherein, The determining the splitting condition of the initial reply content according to the user interaction data includes: Determine the splitting condition of the initial reply content according to the current user interaction data and / or the historical user interaction data.

30. The method of claim 28, wherein, The determining the splitting condition of the initial reply content according to the user interaction data includes: When the user interaction data is first user interaction data, determine that the splitting condition of the initial reply content is splitting according to the first user interaction data; When the user interaction data is second user interaction data, determine that the splitting condition of the initial reply content is not splitting according to the second user interaction data.

31. The method of claim 28, wherein, When the splitting condition of the initial reply content is splitting, the initial reply content includes at least two first reply contents.

32. A reply message generating method characterized by comprising the steps of: The method includes: Obtain user interaction data; Determine a guiding type corresponding to the user interaction data according to the user interaction data; According to the user interaction data and the guidance type, reply information is generated; the reply information at least includes first reply content; whether the reply information includes second reply content corresponds to the guidance type.

33. A reply message generating method characterized by comprising the steps of: The method comprises: Obtaining user interaction data; According to the user interaction data, determine the reply information structure; According to the user interaction data and the reply information structure, generate reply information.

34. A reply information generating apparatus characterized by comprising: Including: Interaction data acquisition module, for obtaining user interaction data; Reply information generation module, for generating reply information according to the user interaction data, the reply information at least includes first reply content, the information amount of the first reply content corresponds to the user interaction data.

35. An electronic device, comprising: The electronic device comprises: At least one processor; and The memory is in communication connection with the at least one processor; wherein The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the reply information generation method in any one of claims 1-31.

36. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to enable the processor to execute the reply information generation method in any one of claims 1-31 when executed.

Images