Interaction method and device based on large model, related equipment and program product

Abstract

The invention discloses an interaction method and device based on a large model, related equipment and a program product, and the method comprises the steps: determining the speaker role information of a current interaction request through recognizing a current interaction text corresponding to the voice of the current interaction request; in order to prevent the large model from responding to invalid requests such as everyone dialogue and noise, calling the large model, indicating the large model to determine dialogue state information of the current-round interaction request in combination with information such as total interaction number, speaker role information and current-round and historical interaction texts, and displaying the dialogue state information; the information comprises output indication information representing whether the large model is indicated by a user to answer or not and topic attribute information, judging whether the large model of the current interaction request needs to respond or not in combination with dialogue state information, and calling the large model to generate a dialogue reply of the current interaction request and outputting the dialogue reply when the dialogue reply needs to respond. According to the scheme, invalid responses to data such as conversation and noise can be reduced, and the overall intelligence and interaction fluency of man-machine interaction are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to an interaction method, apparatus, related equipment and program product based on a large model. Background Art

[0002] With the development of general large model technology, the effect of human-computer dialogue has been greatly improved. In the current full-duplex interactive system, the speaker's voice content is recognized into text through speech recognition technology and then sent to the large model. The large model will respond to all requests, including three categories: human-computer dialogue, human-to-human dialogue, and noise. The large model will also give invalid responses to human-to-human dialogue and noise data, and interrupt the user's normal human-to-human conversation, seriously affecting the user's normal interaction. Summary of the invention

[0003] In view of the above problems, this application is proposed to provide an interactive method, device, related equipment and program product based on a large model to enhance the intelligence of the large model in a multi-person chat scenario and improve the user's interactive experience. The specific solution is as follows:

[0004] In a first aspect, a large model-based interactive method is provided, comprising:

[0005] Identify the current round of interaction text corresponding to the voice of the current round of interaction request, and determine the speaker role information of the current round of interaction request;

[0006] Calling the big model to instruct the big model to determine the dialogue state information of this round of interaction request in combination with input information, the input information includes the total number of interactors, the speaker role information, the text of this round of interaction and the historical interaction text, and the dialogue state information includes output indication information indicating whether the big model itself is instructed by the user to answer, and topic attribute information;

[0007] At least in combination with the dialogue state information, determine whether the big model of this round of interaction request needs to respond, and when a response is required, call the big model to generate a dialogue response for this round of interaction request, and output the dialogue response.

[0008] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the method further includes:

[0009] Determine the speaker attribute information of this round of interaction request;

[0010] Then the input information also includes the speaker attribute information.

[0011] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the process of determining the speaker role information and the speaker attribute information of the current round of interaction request includes:

[0012] Perform voiceprint recognition on the voice of this round of interaction request and compare it with the set voiceprints of each role to obtain the role information of the speaker of this round of interaction request;

[0013] Determine the attribute information of the speaker of this round of interaction request through the voiceprint recognition result, wherein the attribute information includes at least one of the following: gender and age of the speaker;

[0014] Or, query the configured speaker attribute information corresponding to the speaker role of the current round of interaction request.

[0015] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the method further includes:

[0016] Get the interaction interval time between the current round of interaction requests and the previous round of interaction requests;

[0017] Then, the process of determining whether the current round of interaction request big model needs to respond, at least in combination with the dialog state information, includes:

[0018] In combination with the dialog state information and the interaction interval time, it is determined whether the large model of this round of interaction request needs to respond.

[0019] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the process of determining whether the current round of interaction request big model needs to respond, at least in combination with the dialogue state information, includes:

[0020] If the output instruction information indicates that the large model itself is instructed by the user to answer, it is determined that the large model needs to respond to the current round of interaction request.

[0021] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type;

[0022] The process of determining whether the current round of interaction request big model needs to respond at least in combination with the dialog state information further includes:

[0023] Based on the results of each item in the topic attribute information, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond;

[0024] The large model response condition is composed of one or more value results of topic attribute information configured by the user.

[0025] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type; the dialogue state information also includes the user emotion of the current round of interaction request;

[0026] The process of determining whether the current round of interaction request big model needs to respond at least in combination with the dialog state information further includes:

[0027] Based on the results of each item in the topic attribute information and the user emotion, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond;

[0028] The large model response condition is configured by the user and is composed of one or more value results of topic attribute information and / or user emotion type.

[0029] In a possible design, in another implementation of the first aspect of the embodiments of the present application, the large model response condition includes at least one of the following:

[0030] The topic is complete, there is continuity in the topic, and the topic has been concluded;

[0031] The topic is complete and the topic type is to seek a response;

[0032] The topic is complete, the topic type is emotional soothing, and the user's emotion is negative.

[0033] In a possible design, in another implementation of the first aspect of the embodiment of the present application, the process of determining whether the current round of interaction request big model needs to respond in combination with the dialogue state information and the interaction interval time includes:

[0034] Based on the dialog state information, determining whether the current round of interaction request big model needs to respond; and,

[0035] It is determined whether the interaction interval time exceeds a set time threshold, and if so, it is determined that the large model of this round of interaction request needs to respond.

[0036] In a possible design, in another implementation of the first aspect of the embodiment of the present application, when a response is required, the process of calling the large model to generate a dialogue reply to the current round of interaction request includes:

[0037] When it is determined that a response is required because the interaction interval time exceeds a set time threshold, the large model is called to generate a dialogue response for a recommended topic based on the speaker attribute information of the set interaction role and / or the interests shown in the historical interactions;

[0038] When it is determined that a response is required due to other reasons, the large model is called to generate a dialogue response by combining the current round of interaction text and the historical interaction text.

[0039] In a second aspect, an interactive device based on a large model is provided, comprising:

[0040] A speech recognition unit, used to recognize the current round of interaction text corresponding to the speech of the current round of interaction request;

[0041] A role identification unit, used to determine the speaker role information of this round of interaction request;

[0042] A dialogue state identification unit is used to call the big model to instruct the big model to determine the dialogue state information of the current round of interaction request in combination with input information, wherein the input information includes the total number of interactors, the speaker role information, the current round of interaction text and the historical interaction text, and the dialogue state information includes output indication information indicating whether the big model itself is instructed by the user to answer, and topic attribute information;

[0043] A response determination unit, configured to determine whether the current round of interaction request big model needs to respond, at least in combination with the dialogue state information;

[0044] The response output unit is used to call the large model to generate a dialogue response to the current round of interaction request when a response is needed, and output the dialogue response.

[0045] In a third aspect, an electronic device is provided, comprising: a memory and a processor;

[0046] The memory is used to store programs;

[0047] The processor is used to execute the program to implement the various steps of the large model-based interaction method described in any one of the first aspects of the present application.

[0048] In a fourth aspect, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the various steps of the large model-based interaction method described in any one of the first aspects of the present application are implemented.

[0049] In a fifth aspect, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the various steps of the large model-based interaction method described in any one of the aforementioned first aspects of the present application.

[0050] By means of the above technical solution, the present application identifies the current round of interaction text corresponding to the voice of the current round of interaction request for the human-computer interaction scenario, and determines the speaker role information of the current round of interaction request. In order to prevent the big model from responding to invalid requests such as person-to-person dialogue and noise, the present application can call the big model, instructing the big model to combine the total number of interactions, speaker role information, current round and historical interaction text and other information to determine the dialogue state information of the current round of interaction request. The dialogue state information includes output indication information representing whether the big model itself is instructed by the user to answer, and topic attribute information, that is, the big model can analyze whether the current round of interaction request is forced to trigger a response by the user, and the topic attribute information discussed in the current round of interaction request. On this basis, it can be combined with the dialogue state information to determine whether the big model of the current round of interaction request needs to respond. When a response is required, the big model can be called to generate and output the dialogue reply of the current round of interaction request. Obviously, the big model in the present application does not respond to all interaction requests, but refers to the input information to determine the dialogue status information of this round of interaction requests, and judges whether the big model needs to respond in this round. This can reduce invalid responses to data such as human dialogues and noise, improve the overall intelligence and smoothness of human-computer interaction, and enhance the user's interactive experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0052] Figure 1 A schematic diagram of an implementation system architecture of the large model-based interaction method provided in an embodiment of the present application;

[0053] Figure 2 A schematic diagram of a large model-based interactive method flow chart provided in an embodiment of the present application;

[0054] Figure 3 A schematic diagram of the structure of an interactive device based on a large model provided in an embodiment of the present application;

[0055] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] It is understandable that before using the technical solutions disclosed in the embodiments of this application, the type, scope of use, usage scenarios, etc. of the personal information involved in this application should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0058] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present application according to the prompt message.

[0059] This application provides an interactive method based on a large model, which can be applied to Figure 1 The system architecture shown in FIG. 1 may include a terminal 100 and a server 200. The server 200 may include one or more servers ( Figure 1 A server is included as an example for explanation).

[0060] The terminal 100 or the server 200 can be used alone to execute the interactive method based on the large model provided in the embodiment of the present application. In addition, the terminal 100 and the server 200 can also be used in collaboration to execute the interactive method based on the large model provided in the embodiment of the present application.

[0061] Next describe Figure 1 The product form of the mid-terminal 100;

[0062] The terminal 100 in the embodiment of the present application can be a mobile phone, a tablet computer, a vehicle-mounted device, a teaching screen, a wearable device, a conference terminal, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiment of the present application does not impose any restrictions on this.

[0063] An example of a possible application scenario is that in a car chat scenario, the user in the car can chat with the intelligent car system, and the intelligent car system executes the interaction method based on the big model of this application. The intelligent car system can deploy the big model locally in the vehicle, or can access the big model on the server through the local car system of the vehicle.

[0064] Another possible application scenario example is that multiple users can have a chat conversation with an intelligent robot, and the intelligent robot executes the large model-based interaction method of the present application.

[0065] In addition, the present application method can also be applied to some other multi-person chat scenarios, which are not listed here one by one.

[0066] The current interaction scenario is still a question-and-answer interactive chat between humans and machines. In a multi-person chat scenario, the large model will still "respond" to the "instructions" issued by all people in the process, including invalid responses to data such as human conversations and noise, interrupting the user's normal topics and affecting the user's interactive experience.

[0067] In view of the problems existing in the above-mentioned existing interaction schemes, the present application proposes a human-computer interaction scheme based on a large model. The scheme can distinguish the roles of the speaker through external sound receiving equipment and voiceprint recognition technology, and further identify gender, age and other personal attributes of the speaker and send them to the large model together, and at the same time, the interval time between the current round of interaction request and the previous round of interaction request is also sent to the large model as auxiliary information. The large model first determines whether the user of this round of interaction request is forced to trigger a response, the topic completeness, type and other attribute information of this round of interaction request, and the user's emotional state through the role information, personal attribute information, current round of interaction text and historical interaction text of this round of interaction request. In addition, the interval time between this round of interaction request and the previous round of interaction request can also be determined. On this basis, the above information can be combined to determine whether the large model of this round of interaction responds and what kind of response is given, such as directly answering user questions, summarizing the views of each user, and cold transition topics, thereby realizing the intelligent interaction of the large model of multi-person chat scene.

[0068] The present application embodiment provides an interaction method based on a large model, and takes the method applied to a computer device as an example. The computer device may be Figure 1 The terminal 100 or the system consisting of the terminal 100 and the server 200. Figure 2 , the interactive method based on the large model specifically includes the following steps:

[0069] Step S100: Identify the current round of interaction text corresponding to the voice of the current round of interaction request, and determine the speaker role information of the current round of interaction request.

[0070] Specifically, the interactive voice of each user in the chat scene can be collected through an external audio receiving device, and the user voice of the current round of interactive request can be recognized, and the recognized text can be used as the text of the current round of interaction.

[0071] Furthermore, in this step, the speaker role information of this round of interaction request can also be determined, where the role information can be determined by using voiceprint recognition technology to perform voiceprint recognition on the current round of interaction voice and compare it with the set voiceprints of each role to determine the speaker role. The speaker role can be a number of user roles that are pre-set by the user before the interaction, or a new role will be added every time the system finds a new role during the chat. Different speaker roles represent different speakers.

[0072] In some possible implementations, the present application may further determine the speaker attribute information of the current round of interaction request. The speaker attribute information may be used as the speaker's personality attributes, such as age, gender, etc.

[0073] The process of determining the speaker attribute information of this round of interaction request may be to perform voiceprint recognition on the voice of this round of interaction request, and determine the speaker attribute information of this round of interaction request through the voiceprint recognition result. Another possible implementation method is that after determining the role information of this round of interaction request, the speaker attribute information corresponding to the configured speaker role of this round of interaction request may be queried. That is, the present application may pre-configure the speaker attribute information corresponding to different speaker roles, and after determining the role information of this round of interaction request, the speaker attribute information may be determined by querying the configuration information.

[0074] Step S110, calling the big model to instruct the big model to determine the dialogue state information of this round of interaction request in combination with the input information, where the input information includes the total number of interacting persons, speaker role information, this round of interaction text and historical interaction text.

[0075] In this step, by calling the capability of the big model, the big model is instructed to combine the input information to predict the dialogue state information of this round of interaction request, wherein the dialogue state information includes output indication information representing whether the big model itself is instructed by the user to answer, and topic attribute information.

[0076] The input information of the big model includes at least the total number of interactions, that is, the total number of people participating in the chat, the speaker role information of the current round of interaction request, the current round of interaction text and the historical interaction text. The historical interaction text can be the interaction text of all users and the big model within a set time before the current moment (for example, 3 minutes or other values).

[0077] Further optionally, the input information may also include other information, such as speaker attribute information of this round of interaction request.

[0078] The prompt instruction prompt is combined with the above input information and sent to the large model to instruct the large model to generate the dialogue state information of this round of interaction request.

[0079] The dialogue status information contains at least two types of information. The first is the output indication information that characterizes whether the large model itself is instructed by the user to answer. The second is the topic attribute information of this round of interaction request, including but not limited to: whether the topic is complete, whether the topic has continuity, whether the topic is completed, the topic type, etc.

[0080] In some possible implementations, the dialogue state information may further include the user's emotions in this round of interaction request, so that in subsequent steps, the user's emotions can be combined to determine whether the large model needs to intervene in time to appease the user.

[0081] The following is an example of a prompt that calls a large model to generate the dialog status information of this round of interaction request:

[0082] “Suppose you are an intelligent chatbot named Spark. Please use the interaction history to determine the following information about the last round of interaction:

[0083] 1. Determine whether you have been named or explicitly asked to answer a question by the user

[0084] 2. Determine whether the content of this round of dialogue is complete

[0085] 3. Determine whether this round of conversation is a continuous topic

[0086] 4. Determine whether this round of dialogue is completed

[0087] 5. Determine the type of this round of dialogue. The type is defined as follows:

[0088] a. Response seeking type: information seeking type, opinion seeking type, curious exploration type, doubt seeking type

[0089] b. Timely participation: comparison and selection, experience sharing, interest and hobbies, hypothetical scenarios, controversial discussions

[0090] c. Summary and appeasement type: cultural exchange type, experience sharing type, social phenomenon type, natural phenomenon type

[0091] d. Emotional comfort type: humorous and teasing type, emotional expression type

[0092] e. Meaningless

[0093] 6. Determine the emotions expressed by the user during this round of conversation. The definition of emotions is as follows:

[0094] a. Positive: happy, caring, praising, serious, teasing, blessing, coquettish, soft voice

[0095] b. Negative: sad, angry, depressed, confused, surprised

[0096] c. Neutral

[0097] Several examples are provided below:

[0098] Example 1:

[0099] There are 3 people, Role 1 (elderly, male), Role 2 (middle-aged, male), Role 3 (child, female)

[0100] Character 1: Today’s weather is really nice

[0101] Output: ##Forced trigger: No ##Topic complete: Yes ##Topic continuity: No ####Topic completion: Not applicable ##Topic type: Summary and appeasement type - natural phenomenon type ##User emotion: Neutral ##

[0102] Example 2:

[0103] There are 3 people, Role 1 (elderly, male), Role 2 (middle-aged, male), Role 3 (child, female)

[0104] Character 1: It's so annoying. It's getting cold out of nowhere. Everyone, tell me what you want to eat tonight.

[0105] Role 2: Spark, you speak first

[0106] Output: ##Forced trigger: Yes ##Topic complete: Yes ##Topic continuity: Yes ####Topic completed: No ##Topic type: Seeking response - Soliciting opinions ##User sentiment: Negative - Depressed ##

[0107] Example 3:

[0108] There are 3 people, Role 1 (elderly, male), Role 2 (middle-aged, male), Role 3 (child, female)

[0109] Character 1: This weather is really cold

[0110] Role 2: That

[0111] Output: ##Forced trigger: No ##Topic complete: No ##Topic continuity: No ####Topic completion: Not applicable ##Topic type: Not applicable ##User sentiment: Neutral ##

[0112] Interaction History:

[0113] There are 3 people, Role 1 (elderly, male), Role 2 (middle-aged, male), Role 3 (child, female)

[0114] Role 1: It's getting cold today. Let's talk about what we want to eat in the evening.

[0115] Role 2: I want to eat hot pot."

[0116] Corresponding to the prompt in the above example, the output example of the large model is as follows:

[0117] Output: ##Forced trigger: No##Topic complete: Yes##Is the topic continuous: Yes####Topic completed: No##Topic type: Timely participation type - experience sharing type##User sentiment: neutral##.

[0118] Among them, the mandatory trigger item is the output indication information that characterizes whether the large model itself is instructed by the user to answer. When the mandatory trigger item is no, it means that the large model in this round of interaction has not been instructed by the user to answer. The topic completeness item characterizes whether the topic of this round of interaction request is complete. The topic continuity item characterizes whether the topic of this round of interaction request has continuity. The topic completion item characterizes whether this round of interaction is completed. The topic type item characterizes the topic type of this round of interaction request, which can be taken from various types of examples. The user emotion item characterizes the user emotion type of this round of interaction request, and its value can be taken from various types of examples.

[0119] Step S120: at least in combination with the dialogue state information, determine whether the big model of this round of interaction request needs to respond, and when a response is needed, call the big model to generate a dialogue response for this round of interaction request, and output the dialogue response.

[0120] Specifically, the dialog state information output by the large model can be obtained through the previous step. In this step, the dialog state information can be combined to determine whether the large model needs to respond to the current round of interaction request.

[0121] In another possible implementation, the method of the present application can further obtain the interaction interval time between the current round of interaction requests and the previous round of interaction requests. Exemplarily, a system timer can be set to continuously refresh and record the interaction interval time between the system input time of the most recent round of interaction requests and the end time of the previous round of interaction responses. If the interaction interval time exceeds the set time threshold, it indicates that the conversation interval time is long, otherwise, it indicates that the conversation interval time is short.

[0122] On this basis, the dialogue state information and the interaction interval time can be combined to determine whether the large model of this round of interaction request needs to respond.

[0123] The interactive method provided in the embodiment of the present application identifies the current round of interactive text corresponding to the voice of the current round of interactive request for the human-computer interaction scenario, and determines the speaker role information of the current round of interactive request. In order to prevent the big model from responding to invalid requests such as person-to-person dialogue and noise, the present application can call the big model to instruct the big model to determine the dialogue state information of the current round of interactive request in combination with the total number of people interacting, the speaker role information, the current round and historical interactive texts and other information. The dialogue state information includes output indication information representing whether the big model itself is instructed by the user to answer, and topic attribute information, that is, the big model can analyze whether the current round is forced to trigger a response by the user, and the topic attribute information discussed in the current round of interactive request. On this basis, it can be combined with the dialogue state information to determine whether the big model of the current round of interactive request needs to respond. When a response is required, the big model can be called to generate and output the dialogue reply of the current round of interactive request. Obviously, the big model in the present application does not respond to all interaction requests, but refers to the input information to determine the dialogue status information of this round of interaction requests, and judges whether the big model needs to respond in this round. This can reduce invalid responses to data such as human dialogues and noise, improve the overall intelligence and smoothness of human-computer interaction, and enhance the user's interactive experience.

[0124] In some embodiments of the present application, the process of combining the aforementioned step S120 with the dialogue state information to determine whether the current round of interaction request large model needs to respond is described.

[0125] The dialogue status information includes output indication information that characterizes whether the large model itself is instructed by the user to answer. Therefore, if the output indication information indicates that the large model itself is instructed by the user to answer, that is, the user forcibly triggers the large model to respond, it can be directly determined that the large model needs to respond to this round of interaction request.

[0126] Furthermore, the conversation state information also includes topic attribute information, including but not limited to at least one of the following: whether the topic is complete, whether the topic is continuous, whether the topic is finished, the topic type, etc. On this basis, in this embodiment, it is also possible to determine whether the configured big model response conditions are met based on the results of each item in the topic attribute information. If any big model response condition is met, it is determined that the big model of this round of interaction request needs to respond.

[0127] Among them, the large model response condition is composed of the value results of one or more topic attribute information configured by the user.

[0128] The following are some examples of optional large model response conditions, including but not limited to:

[0129] Condition 1: The topic is complete, has continuity, and has been concluded.

[0130] When condition 1 is met, it means that the topic discussed in this round of interaction request has been completed, and the big model can summarize the topic to improve the participation of the big model in multi-person chat scenarios.

[0131] Condition 2: The topic is complete and the topic type is seeking a response.

[0132] Among them, the topic type of seeking a reply indicates that the current topic is that the user is seeking a reply from the big model, and the big model can respond under the same conditions.

[0133] Among them, the response-seeking type includes but is not limited to: information-seeking type, opinion-seeking type, curiosity-exploring type, doubt-solving type, etc., which can be specifically set by the user.

[0134] The above only illustrates several optional examples of large model response conditions. In addition, users can also configure large model response conditions composed of the value results of one or more other topic attribute information according to actual needs, which are not listed here one by one.

[0135] In some possible implementations, the dialogue state information may further include the user emotion of the current round of interaction request. Then step S120 combines the dialogue state information to determine whether the current round of interaction request big model needs to respond, which may include:

[0136] Based on the results of each item in the topic attribute information and the user's emotions, it is determined whether the configured big model response conditions are met. If any big model response condition is met, it is determined that the big model of this round of interaction request needs to respond.

[0137] Among them, the large model response condition is composed of one or more value results of topic attribute information and / or user emotion type configured by the user.

[0138] Based on the two large model response conditions illustrated in the above embodiments, this embodiment further illustrates a third large model response condition, as shown below:

[0139] Condition 3: The topic is complete, the topic type is emotional soothing, and the user's emotion is negative.

[0140] When the topic type of this round of interaction request is emotional soothing, it means that the user is expressing his or her specific emotions. At this time, if the user's emotions are negative, the big model can intervene in time to soothe the user's emotions and adjust the chat atmosphere. In other words, the big model can respond under this condition.

[0141] In other embodiments of the present application, when determining whether the current round of interaction request large model needs to respond, in addition to combining the dialogue state information, the interaction interval time (ie, the interval time between the current round of interaction request and the previous round of interaction request) may also be combined.

[0142] Specifically, on the one hand, based on the dialogue state information, it can be determined whether the current round of interactive request big model needs to respond. The detailed implementation process can refer to the above embodiment, which will not be repeated here.

[0143] On the other hand, it can be determined whether the interaction interval exceeds the set time threshold. If so, it can be determined that the interaction interval is too long. At this time, in order to avoid a dull moment, it can be determined that the big model needs to respond to this round of interaction requests. The big model can intervene in time to respond and throw out a transition topic to improve the fluency of the entire human-computer interaction.

[0144] In some embodiments of the present application, the process of calling the large model to generate a dialogue response to the current round of interaction request when it is determined that a response is required in the aforementioned step S120 is described.

[0145] As described in the above embodiments, when determining whether a large model needs to respond, it can be determined from two aspects: on the one hand, whether the dialogue state information meets the set large model response conditions; on the other hand, whether the interaction interval time exceeds the set time threshold.

[0146] (1) If a response is required because the interaction interval exceeds a set time threshold, the large model can be called to generate a dialogue response for the recommended topic based on the speaker attribute information of the set interaction role and / or the interests shown in historical interactions.

[0147] The set interactive role may be a pre-specified main role, or may be a role that occupies a dominant position in historical interactions, such as a role that speaks the most, etc. The set interactive role may also be a randomly selected role, etc.

[0148] By instructing the large model to generate transitional conversation responses related to the recommended topic based on the speaker attribute information of the set interaction role and / or the interests shown in historical interactions, the chatting users can be guided to continue the subsequent interaction and avoid awkward silence.

[0149] (2) If a response is required for other reasons (such as when the conversation status information meets the set big model response conditions), the big model can be called to generate a conversation response based on the current round of interaction text and historical interaction text.

[0150] Specifically, if the big model needs to respond not because the interaction interval is too long, it means that the big model needs to respond to the specific interaction content. Therefore, the big model can be called to combine the current round of interaction text and the historical interaction text to generate a dialogue response to the current round of interaction request, thereby completing the human-computer dialogue process.

[0151] In this embodiment, when a large model determined for different reasons needs to respond, different strategies are used to generate dialogue replies to better meet different scenario requirements and improve the quality of dialogue replies.

[0152] The following describes an interactive device based on a large model provided in an embodiment of the present application. The interactive device based on a large model described below and the interactive method based on a large model described above can refer to each other.

[0153] See also Figure 3 , Figure 3 A schematic diagram of the structure of an interactive device based on a large model disclosed in an embodiment of the present application.

[0154] like Figure 3 As shown, the device may include:

[0155] A speech recognition unit 11, used to recognize the current round of interaction text corresponding to the speech of the current round of interaction request;

[0156] A role identification unit 12, used to determine the role information of the speaker in this round of interaction request;

[0157] The dialogue state identification unit 13 is used to call the big model to instruct the big model to determine the dialogue state information of the current round of interaction request in combination with the input information, wherein the input information includes the total number of interactors, the speaker role information, the current round of interaction text and the historical interaction text, and the dialogue state information includes output indication information indicating whether the big model itself is instructed by the user to answer, and topic attribute information;

[0158] A response determination unit 14, configured to determine whether the current round of interaction request large model needs to respond, at least in combination with the dialogue state information;

[0159] The response output unit 15 is used to call the large model to generate a dialogue response to the current round of interaction request when a response is needed, and output the dialogue response.

[0160] In a possible implementation, the role identification unit is further used for:

[0161] The speaker attribute information of the current round of interaction request is determined, and on this basis, the input information also includes the speaker attribute information.

[0162] In a possible implementation, the process of the role identification unit determining the speaker role information and speaker attribute information of the current round of interaction request includes:

[0163] Perform voiceprint recognition on the voice of this round of interaction request and compare it with the set voiceprints of each role to obtain the role information of the speaker of this round of interaction request;

[0164] Determine the attribute information of the speaker of this round of interaction request through the voiceprint recognition result, wherein the attribute information includes at least one of the following: gender and age of the speaker;

[0165] Or, query the configured speaker attribute information corresponding to the speaker role of the current round of interaction request.

[0166] In a possible implementation, the device of the present application may further include:

[0167] An interaction interval determination unit, used to obtain the interaction interval between the current round of interaction requests and the previous round of interaction requests;

[0168] The process in which the response determination unit determines whether the current round of interaction request big model needs to respond based on at least the dialog state information includes:

[0169] In combination with the dialog state information and the interaction interval time, it is determined whether the large model of this round of interaction request needs to respond.

[0170] In a possible implementation, the process in which the response determination unit determines whether the current round of interaction request large model needs to respond by at least combining the dialog state information includes:

[0171] If the output instruction information indicates that the large model itself is instructed by the user to answer, it is determined that the large model needs to respond to the current round of interaction request.

[0172] In a possible implementation, the topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type; the response determination unit determines whether the current round of interaction request big model needs to respond based on at least the dialogue state information, and further includes:

[0173] Based on the results of each item in the topic attribute information, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond;

[0174] The large model response condition is composed of one or more value results of topic attribute information configured by the user.

[0175] In a possible implementation, the topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type; the dialogue state information also includes the user emotion of the current round of interaction request; the response determination unit determines whether the current round of interaction request big model needs to respond based on at least the dialogue state information, and the process also includes:

[0176] Based on the results of each item in the topic attribute information and the user emotion, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond;

[0177] The large model response condition is configured by the user and is composed of one or more value results of topic attribute information and / or user emotion type.

[0178] In a possible implementation, the large model response condition includes at least one of the following:

[0179] The topic is complete, there is continuity in the topic, and the topic has been concluded;

[0180] The topic is complete and the topic type is to seek a response;

[0181] The topic is complete, the topic type is emotional soothing, and the user's emotion is negative.

[0182] In a possible implementation, the process in which the response determination unit determines whether the current round of interaction request big model needs to respond based on the dialog state information and the interaction interval time includes:

[0183] Based on the dialog state information, determining whether the current round of interaction request big model needs to respond; and,

[0184] It is determined whether the interaction interval time exceeds a set time threshold, and if so, it is determined that the large model of this round of interaction request needs to respond.

[0185] In a possible implementation, when a response is required, the response output unit calls the large model to generate a dialog response to the current round of interaction request, including:

[0186] When it is determined that a response is required because the interaction interval time exceeds a set time threshold, the large model is called to generate a dialogue response for a recommended topic based on the speaker attribute information of the set interaction role and / or the interests shown in the historical interactions;

[0187] When it is determined that a response is required due to other reasons, the large model is called to generate a dialogue response by combining the current round of interaction text and the historical interaction text.

[0188] The present application also provides an electronic device in an embodiment. Figure 4 As shown, it shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present application. The electronic device in the embodiment of the present application may include but is not limited to fixed terminals such as mobile phones, tablet computers, vehicle-mounted terminals, wearable devices, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0189] like Figure 4 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 to a random access memory (RAM) 603, so as to implement the large model-based interaction method of the aforementioned embodiment of the present application. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0190] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0191] Also provided in an embodiment of the present application is a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any large model-based interaction method provided in an embodiment of the present application.

[0192] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any large model-based interaction method provided in the embodiment of the present application.

[0193] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0194] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0195] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0196] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0197] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

Claims

1. An interactive method based on a large model, characterized in that: include: Identify the current round of interaction text corresponding to the voice of the current round of interaction request, and determine the speaker role information of the current round of interaction request; Calling the big model to instruct the big model to determine the dialogue state information of this round of interaction request in combination with input information, the input information includes the total number of interactors, the speaker role information, the text of this round of interaction and the historical interaction text, and the dialogue state information includes output indication information indicating whether the big model itself is instructed by the user to answer, and topic attribute information; At least in combination with the dialogue state information, determine whether the big model of this round of interaction request needs to respond, and when a response is required, call the big model to generate a dialogue response for this round of interaction request, and output the dialogue response.

2. The method according to claim 1, characterized in that Also includes: Determine the speaker attribute information of this round of interaction request; Then the input information also includes the speaker attribute information.

3. The method according to claim 2, characterized in that The process of determining the speaker role information and speaker attribute information of this round of interaction request includes: Perform voiceprint recognition on the voice of this round of interaction request and compare it with the set voiceprints of each role to obtain the role information of the speaker of this round of interaction request; Determine the attribute information of the speaker of this round of interaction request through the voiceprint recognition result, wherein the attribute information includes at least one of the following: gender and age of the speaker; Or, query the configured speaker attribute information corresponding to the speaker role of the current round of interaction request.

4. The method according to claim 1, characterized in that: Also includes: Get the interaction interval time between the current round of interaction requests and the previous round of interaction requests; Then, the process of determining whether the current round of interaction request big model needs to respond, at least in combination with the dialog state information, includes: In combination with the dialog state information and the interaction interval time, it is determined whether the large model of this round of interaction request needs to respond.

5. The method according to claim 1, characterized in that The process of determining whether the current round of interaction request big model needs to respond at least in combination with the dialog state information includes: If the output instruction information indicates that the large model itself is instructed by the user to answer, it is determined that the large model needs to respond to the current round of interaction request.

6. The method according to claim 5, characterized in that The topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type; The process of determining whether the current round of interaction request big model needs to respond at least in combination with the dialog state information further includes: Based on the results of each item in the topic attribute information, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond; The large model response condition is composed of one or more value results of topic attribute information configured by the user.

7. The method according to claim 5, characterized in that The topic attribute information includes at least one of the following: whether the topic is complete, whether the topic has continuity, whether the topic is finished, and the topic type; the dialogue state information also includes the user emotion of the current round of interaction request; The process of determining whether the current round of interaction request big model needs to respond at least in combination with the dialog state information further includes: Based on the results of each item in the topic attribute information and the user emotion, determine whether the configured big model response conditions are met, and if any big model response condition is met, determine that the big model of this round of interaction request needs to respond; The large model response condition is configured by the user and is composed of one or more value results of topic attribute information and / or user emotion type.

8. The method according to claim 7, characterized in that The large model response condition includes at least one of the following: The topic is complete, there is continuity in the topic, and the topic has been concluded; The topic is complete and the topic type is to seek a response; The topic is complete, the topic type is emotional soothing, and the user's emotion is negative.

9. The method according to claim 4, characterized in that The process of determining whether the current round of interaction request big model needs to respond in combination with the dialog state information and the interaction interval time includes: Based on the dialog state information, determining whether the current round of interaction request big model needs to respond; and, It is determined whether the interaction interval time exceeds a set time threshold, and if so, it is determined that the large model of this round of interaction request needs to respond.

10. The method according to claim 9, characterized in that When a response is required, the process of calling the large model to generate a dialogue response for this round of interaction request includes: When it is determined that a response is required because the interaction interval time exceeds a set time threshold, the large model is called to generate a dialogue response for a recommended topic based on the speaker attribute information of the set interaction role and / or the interests shown in the historical interactions; When it is determined that a response is required due to other reasons, the large model is called to generate a dialogue response by combining the current round of interaction text and the historical interaction text.

11. An interactive device based on a large model, characterized in that: include: A speech recognition unit, used to recognize the current round of interaction text corresponding to the speech of the current round of interaction request; A role identification unit, used to determine the speaker role information of this round of interaction request; A dialogue state identification unit is used to call the big model to instruct the big model to determine the dialogue state information of the current round of interaction request in combination with input information, wherein the input information includes the total number of interactors, the speaker role information, the current round of interaction text and the historical interaction text, and the dialogue state information includes output indication information indicating whether the big model itself is instructed by the user to answer, and topic attribute information; A response determination unit, configured to determine whether the current round of interaction request big model needs to respond, at least in combination with the dialogue state information; The response output unit is used to call the large model to generate a dialogue response to the current round of interaction request when a response is needed, and output the dialogue response.

12. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement the various steps of the large model-based interaction method as described in any one of claims 1 to 10.

13. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the large model-based interaction method as described in any one of claims 1 to 10 is implemented.

14. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the large model-based interaction method as described in any one of claims 1 to 10 is implemented.

Images