Human-computer interaction method and device, storage medium and program product

Abstract

The invention relates to the technical field of intelligent cabins, in particular to a man-machine interaction method and device, a storage medium and a program product. The method comprises the steps that in response to an obtained target session message, a to-be-queried object corresponding to the target session message is determined, and the state of the to-be-queried object is variable; acquiring a current state of the to-be-queried object; and outputting a target reply message corresponding to the target session message through a preset model according to the current state and the to-be-queried object. Thus, man-machine interaction can be achieved through the preset model, the current state of the to-be-queried object with the variable state is obtained, and comprehensiveness and real-time performance of information obtaining by a user are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a human-computer interaction method, device, storage medium, and program product. Background Art

[0002] With the development of artificial intelligence (AI), generative language models such as ChatGPT have gradually emerged, significantly driving the advancement of large model technology. These large models, equipped with a rich knowledge base, can generate text based on a given prompt or question. They can handle tasks with objective answers, such as continuous conversation and question-and-answering, as well as tasks without objective answers, such as copywriting. Compared to the fine-tuned model technology previously employed in the industry for specific domain queries, the Large Language Model (LLM) combined with RAG knowledge-based prompt engineering technology not only achieves better business results but also simplifies practical application. Currently, large models are primarily used in scenarios such as question-and-answering, video, and text creation.

[0003] In existing scenarios where large models are used for human-computer interaction, answers are typically obtained by extracting textual knowledge corresponding to user questions from a static knowledge base. Static knowledge bases store common questions, fixed answers, and related information. However, this model has significant limitations: users can only access static textual knowledge corresponding to their questions, not dynamic textual knowledge, making it difficult to meet users' needs for real-time, dynamic information. Summary of the Invention

[0004] To overcome the problems existing in the related art, the present disclosure provides a human-computer interaction method, device, storage medium and program product.

[0005] According to a first aspect of an embodiment of the present disclosure, a human-computer interaction method is provided, comprising: In response to obtaining the target conversation message, determining an object to be queried corresponding to the target conversation message, where the state of the object to be queried is changeable; Obtain the current state of the object to be queried; According to the current state and the object to be queried, a target reply message corresponding to the target conversation message is output through a preset model.

[0006] Optionally, the object to be queried is an object related to the target device; and obtaining the current state of the object to be queried includes: Sending a request to the target device for instructing to obtain the current status of the object to be queried; receiving identification information of the current state fed back by the target device in response to the request; Determining the current state corresponding to the identification information in a preset configuration information library; The configuration information library includes at least one state and identification information corresponding to the state.

[0007] Optionally, the request includes identification information of the object to be queried, and obtaining the current status of the object to be queried further includes: Determining identification information of the object to be queried in the configuration information database; The configuration information library also includes at least one object that can be queried and identification information corresponding to each of the objects.

[0008] Optionally, in response to obtaining the target conversation message, determining the object to be queried corresponding to the target conversation message includes: In response to obtaining a target conversation message input by a user, performing vectorization processing on a first question included in the target conversation message to obtain a first conversation vector; determining a target session vector according to the first session vector; The to-be-queried object corresponding to the target conversation vector is determined as the to-be-queried object corresponding to the target conversation message.

[0009] Optionally, determining a target session vector according to the first session vector includes at least one of the following: determining a target conversation vector that matches the first conversation vector from at least one second conversation vector included in a pre-built conversation vector library, wherein the second conversation vector is determined based on an object that can be queried and a preset model; or The first session vector is determined as a target session vector.

[0010] Optionally, the conversation vector library is obtained by determining, for each object that can be queried, a second question corresponding to the object, performing vectorization processing on the second question to obtain a second conversation vector, and associatively storing the object and the second conversation vector.

[0011] Optionally, in response to obtaining the target conversation message, determining the object to be queried corresponding to the target conversation message includes: In response to acquiring the target conversation message, an object to be queried corresponding to the target conversation message is determined in a pre-built conversation message library, where the conversation message library includes at least one object that can be queried and a conversation message corresponding to the object.

[0012] Optionally, the conversation message library includes a conversation vector library, wherein the conversation vector library stores at least one object that can be queried and a conversation vector corresponding to the object in an associated manner; and determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library includes: determining a target session vector according to the target session message; The object corresponding to the target conversation vector in the conversation vector library is determined as the object to be queried corresponding to the target conversation message.

[0013] Optionally, the conversation message library further includes conversation keywords corresponding to the respective objects; and determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library includes: determining a target conversation keyword of the target conversation message; Determining a first conversation keyword that matches the target conversation keyword in the conversation message library; The object corresponding to the first conversation keyword is determined as the to-be-queried object corresponding to the target conversation message.

[0014] Optionally, outputting a target reply message corresponding to the target conversation message through a preset model according to the current state and the object to be queried includes: Determine the text of the current state and the text of the object to be queried as model input text; Inputting the model input text into the preset model to obtain a target reply message corresponding to the target conversation message; The target reply message is displayed.

[0015] Optionally, determining the text of the current state and the text of the object to be queried as model input text includes: Obtaining state switching guidance text related to the object to be queried; The state switching guide text, the text of the current state and the text of the object to be queried are determined as model input text.

[0016] According to a second aspect of an embodiment of the present disclosure, a human-computer interaction device is provided, which is used to implement the steps of the human-computer interaction method as described in the first aspect of the embodiment of the present disclosure.

[0017] According to a third aspect of an embodiment of the present disclosure, there is provided an apparatus for implementing human-computer interaction, including: processor; a memory for storing processor-executable instructions; The processor is configured to execute the instructions to enable the device to perform the steps of the human-computer interaction method as described in the first aspect of the embodiment of the present disclosure.

[0018] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the human-computer interaction method provided by the first aspect of the embodiment of the present disclosure are implemented.

[0019] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the human-computer interaction method described in the first aspect of an embodiment of the present disclosure.

[0020] Using this technical solution, in response to receiving a target conversation message, the query object and its current state are determined. Based on the current state and the query object, a target reply message corresponding to the target conversation message is output using a preset model. This enables human-computer interaction through a preset model, allowing the current state of a variable-state query object to be obtained, improving the comprehensiveness and real-time nature of information access for users.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0023] Figure 1 The figure is a flowchart of a human-computer interaction method according to an exemplary embodiment.

[0024] Figure 2 The figure is a schematic diagram showing a method of creating a session vector library according to an exemplary embodiment.

[0025] Figure 3 It is a schematic diagram showing a human-computer interaction according to an exemplary embodiment.

[0026] Figure 4 The figure is a schematic diagram showing a method of creating a static knowledge base according to an exemplary embodiment.

[0027] Figure 5 The present invention is a human-computer interaction device according to an exemplary embodiment.

[0028] Figure 6 is a block diagram of a vehicle according to an exemplary embodiment.

[0029] Figure 7 The figure is a block diagram of a server according to an exemplary embodiment. DETAILED DESCRIPTION

[0030] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0031] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the corresponding data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0032] In the first implementation scenario, the human-computer interaction method provided by the present disclosure is applicable to vehicle scenarios. Users can obtain real-time dynamic signals from the vehicle computer through the human-computer interaction method. For example, they can obtain the vehicle's current gear information, remaining energy storage, wiper working status, window switch status, etc.

[0033] In the second implementation scenario, the human-computer interaction method provided by the present disclosure is applicable to the mobile terminal scenario. Users can obtain real-time dynamic signals of the mobile terminal through the human-computer interaction method. For example, they can obtain the current remaining power of the mobile terminal, the current communication quality, the currently running application, etc. in real time.

[0034] It should be understood that in the first implementation scenario described above, the human-computer interaction method provided by the present disclosure can be executed by the vehicle side, by a server on the vehicle side, or jointly by the vehicle side and the server on the vehicle side. Similarly, in the second implementation scenario described above, the human-computer interaction method provided by the present disclosure can be executed by the mobile terminal, by a server on the mobile terminal side, or jointly by the mobile terminal and the server on the mobile terminal side.

[0035] Figure 1 FIG. 1 is a flow chart of a human-computer interaction method according to an exemplary embodiment. Figure 1 As shown, the human-computer interaction method may include the following steps.

[0036] In step S11 , in response to obtaining the target conversation message, the object to be queried corresponding to the target conversation message is determined, and the state of the object to be queried is changeable.

[0037] In the present disclosure, a target conversation message indicates a question input by the user. For example, the question input by the user may be an inquiry about the current state of the query object. Furthermore, the query object is an object whose state can change rather than remain fixed. For example, in a vehicle scenario, the query object may include the vehicle's current gear information, remaining energy level, wiper operating status, window switch status, and so on.

[0038] The object to be queried corresponding to the target conversation message may be determined based on keywords in the question indicated by the target conversation message, or may be determined based on semantic information of the question indicated by the target conversation message, which is not limited in this disclosure.

[0039] In one embodiment, the target conversation message may be delivered through voice, gesture, text conversation, etc.

[0040] In step S12, the current state of the object to be queried is obtained.

[0041] The variable state of the object to be queried means that the state of the object to be queried is not fixed. Each time the user queries the state of the object, the current state of the object must be obtained from the bottom layer of the device. In the vehicle scenario, the object to be queried is an object related to the vehicle. For example, the object to be queried is the position of the passenger window. Since the position of the passenger window is variable. For example, sometimes the position of the passenger window is completely closed, sometimes it is completely open, and sometimes it is open or closed by a preset ratio, etc. Therefore, each time the object is queried, the current state of the passenger window must be obtained from the bottom layer of the vehicle computer.

[0042] In step S13, according to the current state and the object to be queried, a target reply message corresponding to the target conversation message is output through a preset model.

[0043] After determining the current state of the subject to be queried, the current state and the subject to be queried are input into a preset model to obtain a target reply message corresponding to the target conversation message output by the preset model. Furthermore, after obtaining the target reply message, the target reply message can be displayed on a display device for easy viewing by the user, and / or the target reply message can be played back via voice.

[0044] Alternatively, the preset model can be a large language model that can be refined based on the current state of the input and the query object to generate a target response message. For example, assuming the query object is driving mode, and the current state of the driving mode is sport mode, then "driving mode" and "sport mode" are input into the preset model, and the model outputs a target response message. The target response message may include the user's question "What is the current driving mode?" and the corresponding response text "The current driving mode is sport mode."

[0045] Using this technical solution, in response to receiving a target conversation message, the query object and its current state are determined. Based on the current state and the query object, a target reply message corresponding to the target conversation message is output using a preset model. This enables human-computer interaction through a preset model, allowing the current state of a state-variable query object to be obtained, improving the comprehensiveness and real-time nature of information access for users.

[0046] The following describes a method for determining the object to be queried corresponding to the target conversation message.

[0047] In one embodiment, in response to obtaining a target conversation message, determining the to-be-queried object corresponding to the target conversation message may include: In response to obtaining a target conversation message input by the user, vectorizing a first question included in the target conversation message to obtain a first conversation vector; determining a target session vector according to the first session vector; The to-be-queried object corresponding to the target conversation vector is determined as the to-be-queried object corresponding to the target conversation message.

[0048] The target conversation message may include a first question input by the user, which may be input via voice, text, gesture, or other means. After receiving the first question input by the user, the device executing the human-computer interaction method may vectorize the first question using a trained vector model to obtain a first conversation vector, and then determine a target conversation vector based on the first conversation vector. The training method for the vector model is a relatively mature technology and is not limited in this disclosure.

[0049] In this embodiment, considering the large number of user-vehicle related objects, relying solely on traditional intent understanding to identify the user's query intent based on the first question input would require the vehicle or server to support semantic understanding of thousands of signals, which is difficult to achieve. For example, similar objects such as small mileage A and small mileage B pose a significant challenge to the model's slot extraction.

[0050] Therefore, in one embodiment, determining a target session vector based on a first session vector may include: determining a target session vector that matches the first session vector from at least one second session vector included in a pre-built session vector library, the second session vector being determined based on an object that can be queried and a preset model.

[0051] The conversation vector library is obtained by determining the second question corresponding to each object that can be queried, vectorizing the second question to obtain a second conversation vector, and associating and storing the object with the second conversation vector.

[0052] For example, a session vector library can be constructed as follows: Get multiple objects that can be queried; For each object, a second question corresponding to the object is determined, the second question is vectorized to obtain a second conversation vector, and the object and the second conversation vector are associated and stored to obtain a conversation vector library.

[0053] First, signals are collected to obtain objects that can be queried. For example, in a vehicle scenario, multi-source signals are collected to obtain the original signal. For example, multi-source signals can include but are not limited to: network communication signals, control signals based on the underlying operating system (such as system control signals), and signals related to IoT devices. IoT devices include but are not limited to: fragrance machines, refrigerators, Bluetooth speakers, Bluetooth headsets, etc.

[0054] Different methods can be used to collect signals from different sources. For example, Figure 2 If the object is a network communication signal, it can be obtained from the network signal platform through an API request. If the object is a system control signal, it can be parsed from the context reported by the vehicle computer. Similarly, if the object is a signal related to an IoT device, it can also be parsed from the context reported by the vehicle computer.

[0055] In one embodiment, after the original signal is collected, the original signal can be directly determined as an object that can be queried.

[0056] In another embodiment, after collecting the original signal, the original signal may be pre-processed to obtain the queryable object. Pre-processing the original signal refers to cleaning the original signal, which may include but is not limited to removing duplicate or invalid signals.

[0057] For example, refer to Figure 2 When the target is network communication signals, since network communication signals are difficult for users to understand, a large language model can be used to preprocess the acquired network communication signals to clean them. When the target is system control signals, since there are fewer system control signals, they can be cleaned through manual combing. Similarly, when the target is signals related to IoT devices, manual combing can be used to clean the signals related to IoT devices.

[0058] The pre-processed objects (i.e., signals) can be combined into a full signal set. Then, a corresponding second question is determined for each object included in the full signal set. For example, a corresponding search text, i.e., a second question, can be generated for each object. For example, Figure 2 As shown, the second question corresponding to each object can be obtained through a large language model.

[0059] For example, assuming the object name is "signal": "XXXXXX", the corresponding Chinese name is "name": "XX track switch status feedback", and the Chinese description of the object is "desc": "Displays the current on or off status of the vehicle's traction control system in XX track mode, used for manual adjustment of the track." The keywords expanded for this object are "alias": ["track traction switch status", "XX anti-skid control status", "XX track mode switch"]. A text or instruction prompt added to the large language model. The large language model can generate the second question corresponding to the object. For example, the second question generated for the object can be "queries": ["Is XX track mode on now?", "Is traction control a track-only state"].

[0060] It should be understood that for each object, the number of corresponding second questions determined may be one or more, and this disclosure does not limit this.

[0061] After obtaining the second question corresponding to each object in the above manner, Figure 2 As shown, the second question can also be vectorized to obtain a second conversation vector, and the storage object and the second conversation vector are associated to obtain a conversation vector library. Thus, the conversation vector library can be constructed.

[0062] First of all, it should be understood that for cold start objects, that is, signals unique to new cars that have no user history of querying, the conversation vector library can be expanded through online continuous operation mining.

[0063] After obtaining a first question input by the user and performing vectorization processing on the first question to obtain a first conversation vector, a target conversation vector matching the first conversation vector is searched in the conversation vector library.

[0064] For example, the similarity between the first conversation vector and each second conversation vector in the conversation vector library can be calculated, and the second conversation vector with a similarity greater than a preset threshold can be determined as a target conversation vector that matches the first conversation vector. Alternatively, after calculating the similarity, the second conversation vectors can be sorted in descending order of similarity, and the first K second conversation vectors can be determined as target conversation vectors that match the first conversation vector, where K can be an integer greater than or equal to 1.

[0065] For example, if the user enters the question "Are the headlights off?", the first question is first converted into a vector to obtain a first conversation vector. Based on the first conversation vector, the second conversation vector corresponding to the question "Are the headlights on or off?" in the conversation vector library is determined as the target conversation vector that matches the first conversation vector. The target conversation vector is then used to search for the corresponding query object.

[0066] In the present disclosure, a target conversation vector that matches the first conversation vector can be searched in the conversation vector library, and the number of target conversation vectors can be one or more. In this way, even if the first question input by the user is unclear, the object to be queried can be determined through the target conversation vector that matches it, thereby improving the user experience.

[0067] In another embodiment, the first session vector can be determined as the target session vector. This simplifies the process of determining the target session vector. Referring to the above embodiment, the object corresponding to the first session vector is searched in a pre-built session vector library, and the object corresponding to the first session vector is determined as the object to be queried.

[0068] In the present disclosure, step S11 determines the object to be queried corresponding to the target conversation message in response to obtaining the target conversation message, and may also include: in response to obtaining the target conversation message, determining the object to be queried corresponding to the target conversation message in a pre-built conversation message library, where the conversation message library includes at least one object that can be queried and a conversation message corresponding to the object.

[0069] For example, the conversation message library includes a plurality of objects that can be queried and conversation messages corresponding to each object.

[0070] In one embodiment, a conversation message includes a question. That is, the target conversation message includes a first question input by the user. Accordingly, the conversation message library may include multiple objects and a second question corresponding to each object. Thus, after obtaining the first question input by the user, the conversation message library is searched for a target question that matches the first question. Thereafter, the object corresponding to the target question is determined as the target query object corresponding to the target conversation message.

[0071] The conversation message library is constructed by obtaining a plurality of objects that can be queried; for each object, determining the second question corresponding to the object based on a large language model, and storing the object and the second question in association to obtain the conversation message library.

[0072] In another embodiment, a target conversation vector can be further derived from the target conversation message, and the target object to be queried can be determined based on the target conversation vector. Accordingly, in this embodiment, the conversation message library can include a conversation vector library, which associates and stores at least one queried object with its corresponding conversation vector. For example, the conversation vector library associates and stores multiple queried objects with their corresponding conversation vectors. The construction of the conversation vector library is described above and will not be further elaborated here.

[0073] In this embodiment, the specific implementation method of determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library is: determining the target conversation vector based on the target conversation message, and determining the object corresponding to the target conversation vector in the conversation vector library as the object to be queried corresponding to the target conversation message.

[0074] The specific implementation of determining the target session vector according to the target session message has been described above and will not be repeated here.

[0075] After determining the target session vector in the above manner, the object corresponding to the target session vector in the session vector library is determined as the object to be queried. For example, if multiple target session vectors are determined, the object corresponding to each target session vector can be determined as the object to be queried. That is, there can be one or more objects to be queried.

[0076] By adopting the above technical solution, the query object corresponding to the target conversation message can be determined through vector retrieval. Since vector retrieval is scalable, it is widely applicable to large-model question-answering systems of other discrete signals, expanding the scope of application of human-computer interaction methods.

[0077] In another embodiment, the conversation message library also includes conversation keywords corresponding to each object. Accordingly, determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library may be implemented by: determining a target conversation keyword for the target conversation message; determining a first conversation keyword matching the target conversation keyword in the conversation message library; and determining the object corresponding to the first conversation keyword as the object to be queried corresponding to the target conversation message.

[0078] In this embodiment, the target conversation message may include a first question input by the user. First, keywords in the first question are determined as target conversation keywords. Next, a first conversation keyword matching the target conversation keyword is determined in the conversation message database. Finally, the object corresponding to the first conversation keyword is determined as the target query object corresponding to the target conversation message.

[0079] By adopting the above technical solution, the object to be queried is determined by keyword matching in a slot manner, which improves the flexibility of determining the object to be queried.

[0080] Given the limited computing performance or power of mobile devices or vehicles, pre-set models are typically deployed on a server, such as the vehicle's server or the mobile device's server. This solves the problem of local model deployment on the device and avoids dependency on the target device's underlying client version. Furthermore, the server's superior computing performance can improve the efficiency of human-computer interaction.

[0081] In one embodiment of the present disclosure, the object to be queried may be an object related to a target device. For example, the target device may be a vehicle or a mobile terminal. Accordingly, obtaining the current state of the object to be queried may include: Sending a request to the target device for obtaining the current status of the object to be queried; receiving identification information of a current state of the target device in response to the request feedback; Determining the current state corresponding to the identification information in a preset configuration information library; The configuration information library includes at least one state and identification information corresponding to the state.

[0082] The current state of the object to be queried transmitted at the bottom layer of the target device is usually expressed in the form of a string rather than a natural text form, that is, the current state received from the target device is expressed in the form of a string, while the state input into the preset model is usually represented in the form of natural text. Therefore, in this embodiment, after receiving the identification information of the current state of the object to be queried from the target device, the identification information needs to be converted into a natural text form.

[0083] The configuration information library can be pre-built and include multiple states. The identification information corresponding to each state matches the identification information for that state in the underlying layer of the target device. For example, if the query object is the passenger window position, and the identification information for the fully closed state of the passenger window position is 0x1, then the identification information for the fully closed state of the passenger window position stored in the configuration information library is also 0x1. Thus, when the identification information 0x1 is received from the target device, the state corresponding to 0x1 can be found in the configuration information library as "fully closed."

[0084] Similarly, in the underlying layer of the target device, the signal is also in the form of a string rather than natural text. Therefore, in this embodiment, the request may include the identification information of the object to be queried, obtain the current status of the object to be queried, and may also include: determining the identification information of the object to be queried in the configuration information library; the configuration information library also includes at least one object that can be queried and the identification information corresponding to each object.

[0085] In this disclosure, the configuration information library includes at least one queried object and its corresponding identification information, and at least one state and its corresponding identification information. When constructing the configuration information library, the data source is the full signal set described above. Table 1 shows the information stored in the configuration information library. The identification information corresponding to each object is consistent with the identification information corresponding to the object in the underlying layer of the target device.

[0086] Table 1

[0087] It should be understood that in Table 1, when the object is the wiper control state, state identifier 1 indicates that the wipers are in the off state and not operating. When the object is the active transmission parking gear, state identifier 1 indicates that the vehicle's current gear position is parking. That is, the same identifier can represent different states for different objects.

[0088] In this example, the Chinese names of objects and states are set to facilitate the preset model's understanding of the object and its current state. The Chinese names of objects and states can be batch-edited using the Prompt command and the Large Language Model (LLM). The quality of the edits can then be manually evaluated.

[0089] In the present disclosure, since the state of the object to be queried is variable, in order to accurately obtain the current state of the object to be queried, the current state of the object to be queried needs to be re-acquired during each interaction, that is, during each interaction, the device executing the method needs to send a request to the target device to indicate the acquisition of the current state of the object to be queried, so as to ensure that the current state of the object to be queried can be obtained.

[0090] After determining the current state of the object to be queried according to any of the above methods, a target reply message corresponding to the target conversation message is output through a preset model according to the current state and the object to be queried.

[0091] Optionally, outputting a target reply message corresponding to the target conversation message through a preset model according to the current state and the object to be queried may include: Determine the current state text and the text of the object to be queried as the model input text; Input the model input text into the preset model to obtain the target reply message corresponding to the target conversation message; Displays the target's reply message.

[0092] For example, the text in the current state may be concatenated with the text of the object to be queried to obtain a concatenated text, and the concatenated text may be determined as the model input text.

[0093] The model input text is constructed by concatenating the object to be queried and its current state in the following format: {object}:{current state}. Each object uses a key-value structure, separated by line breaks. The object's value is a natural text identifier.

[0094] For example, suppose the query objects include the passenger seat automatic control, passenger seat ventilation status, and rear left seat heating status. If the passenger seat automatic control is currently in manual mode, the passenger seat ventilation status is currently off, and the rear left seat heating status is currently off, the resulting model input text is "Passenger seat automatic control: Passenger seat automatic control is now manual; Passenger seat ventilation status: Passenger seat ventilation is currently off; Rear left seat heating status: Rear left seat heating is currently off." This model input text is fed into the preset model to obtain the target response message output by the model, which is then displayed.

[0095] Figure 3 FIG. 1 is a schematic diagram showing a human-computer interaction according to an exemplary embodiment. Figure 3 As shown, first, the device executing the human-computer interaction method receives a first question input by the user and vectorizes the first question to obtain a first conversation vector. Next, a target conversation vector matching the first conversation vector is retrieved from a pre-built conversation vector library. The object corresponding to the target conversation vector is determined in the conversation vector library to be the object to be queried. Next, the code signal parsing module parses the natural text of the object to be queried. Based on the natural text of the object to be queried, the configuration information library is retrieved for identification information of the object to be queried. A request is then sent to the target device, requesting the current status of the object to be queried, including the identification information of the object to be queried. For example, assuming the target device is a vehicle, and the natural text of the object to be queried parsed by the code signal parsing module is "position status of the passenger window," and the position status identifier of the passenger window is 0x5, then a request including the identifier 0x5 is sent via the vehicle client to the vehicle's on-board system, instructing the vehicle to query the position status of the passenger window. The vehicle then sends a message to the server, and the vehicle signal parsing module parses the message to obtain the identifier 0x1. After parsing the identifier 0x1, the server searches the configuration information database for the corresponding natural text. For example, if the passenger window is currently fully closed, the corresponding natural text for identifier 0x1 is "fully closed." Finally, the server concatenates the natural text of the query object and the natural text of the current state to generate the model input text, which is then fed into the preset model to generate the target response message.

[0096] In the present disclosure, considering that the user's intention of inquiring about the current state of an object is usually to switch the state of the object, in order to further improve the user experience, in one embodiment, the text of the current state and the text of the object to be queried are determined as model input text, which may include: obtaining the state switching guide text related to the object to be queried; and determining the state switching guide text, the text of the current state and the text of the object to be queried as model input text.

[0097] Reference Figure 3 , it is also possible to query the static knowledge base for state switching guidance text related to the object to be queried. For example, assuming that the object to be queried is the state of a vehicle gear, the relevant text for gear switching, i.e., the state switching guidance text, can be queried from the static knowledge base. The state switching guidance text may include, but is not limited to: operations for switching between different gears, conditions that must be met for switching between different gears, etc. For example, after obtaining the state switching guidance text, the state switching guidance text, the text of the current state, and the text of the object to be queried can be concatenated to obtain the model input text, and then input into the large language model to obtain the target reply message.

[0098] Among them, Figure 4 As shown, the static knowledge base of the vehicle can be obtained by automatically parsing the vehicle's manual.

[0099] For example, suppose the query object is the vehicle's gear status, the current status is D gear, and the status switching guide text switches between different gears. The target response text can be: What gear is it now? The current gear is D gear, which is the forward gear. The forward gear indicator light on the instrument panel should be on. To understand how to switch between gears, please refer to the following operations: From P gear to D gear: Make sure the vehicle is stationary, step on the brake pedal, and push the shift lever down to the limit. From R gear to D gear: Also with the vehicle stationary and the brake pedal pressed, push the shift lever down to the limit. From N gear to D gear: Make sure the vehicle is stationary, step on the brake pedal, and push the shift lever down to the limit.

[0100] In addition to the reply text obtained through the large language model, an image can also be obtained. That is, the target reply message includes the reply text and an image, and the image shows the switching direction between different gears.

[0101] Based on the same inventive concept, the present disclosure further provides a human-computer interaction device, which is used to implement the steps of the human-computer interaction method provided in the present disclosure.

[0102] For example, Figure 5 The present invention is a human-computer interaction device according to an exemplary embodiment. Figure 5FIG. 1 is a block diagram of a human-computer interaction device according to an exemplary embodiment. Figure 5 As shown, the human-computer interaction device 600 may include: A first determining module 601 is configured to, in response to obtaining a target conversation message, determine a to-be-queried object corresponding to the target conversation message, where the state of the to-be-queried object is changeable; A first acquisition module 602 is configured to acquire the current state of the object to be queried; The output module 603 is configured to output a target reply message corresponding to the target conversation message through a preset model according to the current state and the object to be queried.

[0103] Optionally, the object to be queried is an object related to the target device; and the first acquisition module 602 is configured to: Sending a request to the target device for instructing to obtain the current status of the object to be queried; receiving identification information of the current state fed back by the target device in response to the request; Determining the current state corresponding to the identification information in a preset configuration information library; The configuration information library includes at least one state and identification information corresponding to the state.

[0104] Optionally, the request includes identification information of the object to be queried, and the first obtaining module 602 is further configured to: Determining identification information of the object to be queried in the configuration information database; The configuration information library also includes at least one object that can be queried and identification information corresponding to each of the objects.

[0105] Optionally, the first determining module 601 may include: a first processing submodule configured to, in response to obtaining a target conversation message input by a user, perform vectorization processing on a first question included in the target conversation message to obtain a first conversation vector; a first determining submodule, configured to determine a target session vector according to the first session vector; The second determining submodule is configured to determine the to-be-queried object corresponding to the target conversation vector as the to-be-queried object corresponding to the target conversation message.

[0106] Optionally, the first determining submodule is configured to perform at least one of the following: determining a target conversation vector that matches the first conversation vector from at least one second conversation vector included in a pre-built conversation vector library, wherein the second conversation vector is determined based on an object that can be queried and a preset model; or The first session vector is determined as a target session vector.

[0107] Optionally, the conversation vector library is obtained by determining, for each object that can be queried, a second question corresponding to the object, performing vectorization processing on the second question to obtain a second conversation vector, and associatively storing the object and the second conversation vector.

[0108] Optionally, the first determination module 601 is further configured to, in response to obtaining the target conversation message, determine the object to be queried corresponding to the target conversation message in a pre-built conversation message library, wherein the conversation message library includes at least one object that can be queried and the conversation message corresponding to the object.

[0109] Optionally, the conversation message library includes a conversation vector library, which associates and stores at least one object that can be queried and a conversation vector corresponding to the object; the first determination module 601 is further configured to determine a target conversation vector based on a target conversation message; and determine the object corresponding to the target conversation vector in the conversation vector library as the object to be queried corresponding to the target conversation message.

[0110] Optionally, the conversation message library further includes conversation keywords corresponding to each of the objects; the first determining module 601 is further configured to: In response to acquiring the target conversation message, determining a target conversation keyword in the target conversation message; Determining a first conversation keyword that matches the target conversation keyword in the conversation message library; The object corresponding to the first conversation keyword is determined as the to-be-queried object corresponding to the target conversation message.

[0111] Optionally, the output module 603 is configured to: Determine the text of the current state and the text of the object to be queried as model input text; Inputting the model input text into the preset model to obtain a target reply message corresponding to the target conversation message; The target reply message is displayed.

[0112] Optionally, the output module 603 is configured to: Obtaining state switching guidance text related to the object to be queried; The state switching guide text, the text of the current state and the text of the object to be queried are determined as model input text.

[0113] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0114] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which implement the steps of the human-computer interaction method provided by the present disclosure when the program instructions are executed by a processor.

[0115] The present disclosure also provides a device for implementing human-computer interaction, comprising: a processor; a memory for storing processor-executable instructions; The processor is configured to execute the instructions to enable the device to perform the human-computer interaction method as described in the present disclosure.

[0116] For example, the device for implementing human-computer interaction may be an electronic device, such as a vehicle, or a server of the vehicle.

[0117] Figure 6 7 is a block diagram illustrating a vehicle according to an exemplary embodiment. For example, vehicle 700 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or another type of vehicle. Vehicle 700 may be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0118] Reference Figure 6 Vehicle 700 may include various subsystems, such as an infotainment system 710, a perception system 720, a decision-making control system 730, a drive system 740, and a computing platform 750. Vehicle 700 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and each component of vehicle 700 may be interconnected via wired or wireless means.

[0119] In some embodiments, the infotainment system 710 may include a communication system, an entertainment system, a navigation system, and the like.

[0120] Perception system 720 may include several sensors for sensing information about the environment surrounding vehicle 700. For example, perception system 720 may include a global positioning system (which may be a GPS system, a BeiDou system, or another positioning system), an inertial measurement unit (IMU), a laser radar, a millimeter-wave radar, an ultrasonic radar, and a camera.

[0121] The decision control system 730 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0122] The drive system 740 may include components that provide power to the vehicle 700. In one embodiment, the drive system 740 may include an engine, a power source, a transmission system, and wheels. The engine may be an internal combustion engine, an electric motor, an air compression engine, or a combination thereof. The engine is capable of converting energy provided by the power source into mechanical energy.

[0123] Some or all functions of the vehicle 700 are controlled by a computing platform 750. The computing platform 750 may include at least one processor 751 and a memory 752. The processor 751 may execute instructions 753 stored in the memory 752.

[0124] The processor 751 can be any conventional processor, such as a commercially available CPU. The processor can also include a graphics processor (GPU), a field programmable gate array (FPGA), a system on chip (SOC), an application specific integrated circuit (ASIC), or a combination thereof.

[0125] The memory 752 can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0126] In addition to instructions 753 , memory 752 may also store data, such as road maps, route information, and vehicle location, direction, speed, etc. The data stored in memory 752 may be used by computing platform 750 .

[0127] In the embodiment of the present disclosure, the processor 751 may execute the instruction 753 to complete all or part of the steps of the above-mentioned human-computer interaction method.

[0128] Figure 7 1 is a block diagram of a server according to an exemplary embodiment. For example, the server 1900 can be provided as a server. Figure 7The server 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute the instructions to perform the above-mentioned human-computer interaction method.

[0129] Device 1900 may also include a power supply component 1926 configured to perform power management of device 1900, a wired or wireless network interface 1950 configured to connect device 1900 to a network, and an input / output interface 1958. Device 1900 may operate based on an operating system stored in memory 1932.

[0130] Furthermore, the word "exemplary" is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as "exemplary" is not necessarily to be construed as advantageous over other aspects or designs. Rather, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified or clear from the context, "X applies to A or B" is intended to mean any of the natural inclusive permutations. That is, if X applies to A; X applies to B; or X applies to both A and B, then "X applies to A or B" satisfies any of the aforementioned instances. Furthermore, the articles "a" and "an," as used in this application and the appended claims, are generally understood to mean "one or more," unless otherwise specified or clear from the context to refer to the singular form.

[0131] Likewise, although the present disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. The present disclosure includes all such modifications and variations and is limited only by the scope of the claims. With particular regard to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific function of the described component, even if not structurally equivalent to the disclosed structure. In addition, although particular features of the present disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms "include," "have," "have," "have," or variations thereof are used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term "comprising."

[0132] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

[0133] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

[0134] It should be understood that, unless otherwise specifically noted, the features of the various embodiments of the present disclosure described herein may be combined with each other. As used herein, the term "and / or" includes any one of the relevant listed items and any combination of any two or more thereof; similarly, "at least one of" includes any one of the relevant listed items and any combination of any two or more thereof.

[0135] Although terms such as "first", "second" and "third" may be used herein to describe various components, parts, regions, layers or sections, these components, parts, regions, layers or sections are not limited to these terms. On the contrary, these terms are only used to distinguish one component, part, region, layer or section from another component, part, region, layer or section. Therefore, without departing from the teachings of each example, the first component, part, region, layer or section mentioned in the examples described herein may also be referred to as the second component, part, region, layer or section. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" can explicitly or implicitly include at least one such feature. In the description herein, the meaning of "multiple" is at least two, for example, two, three, etc., unless otherwise clearly and specifically defined.

Claims

1. A human-computer interaction method, characterized in that: include: In response to obtaining the target conversation message, determining an object to be queried corresponding to the target conversation message, where the state of the object to be queried is changeable; Obtain the current state of the object to be queried; According to the current state and the object to be queried, a target reply message corresponding to the target conversation message is output through a preset model.

2. The method according to claim 1, characterized in that The object to be queried is an object related to the target device; and obtaining the current state of the object to be queried includes: Sending a request to the target device for instructing to obtain the current status of the object to be queried; receiving identification information of the current state fed back by the target device in response to the request; Determining the current state corresponding to the identification information in a preset configuration information library; The configuration information library includes at least one state and identification information corresponding to the state.

3. The method according to claim 2, characterized in that The request includes identification information of the object to be queried, and obtaining the current status of the object to be queried further includes: Determining identification information of the object to be queried in the configuration information database; The configuration information library also includes at least one object that can be queried and identification information corresponding to each of the objects.

4. The method according to claim 1, wherein In response to obtaining the target session message, determining the to-be-queried object corresponding to the target session message includes: In response to obtaining a target conversation message input by a user, performing vectorization processing on a first question included in the target conversation message to obtain a first conversation vector; determining a target session vector according to the first session vector; The to-be-queried object corresponding to the target conversation vector is determined as the to-be-queried object corresponding to the target conversation message.

5. The method according to claim 4, characterized in that The determining of the target session vector according to the first session vector includes at least one of the following: determining a target conversation vector that matches the first conversation vector from at least one second conversation vector included in a pre-built conversation vector library, wherein the second conversation vector is determined based on an object that can be queried and a preset model; or The first session vector is determined as a target session vector.

6. The method according to claim 5, characterized in that The conversation vector library is obtained by determining, for each object that can be queried, a second question corresponding to the object, performing vectorization processing on the second question to obtain a second conversation vector, and associatively storing the object and the second conversation vector.

7. The method according to claim 1, characterized in that In response to obtaining the target session message, determining the to-be-queried object corresponding to the target session message includes: In response to acquiring the target conversation message, an object to be queried corresponding to the target conversation message is determined in a pre-built conversation message library, where the conversation message library includes at least one object that can be queried and a conversation message corresponding to the object.

8. The method according to claim 7, characterized in that The conversation message library includes a conversation vector library, wherein the conversation vector library stores at least one object that can be queried and a conversation vector corresponding to the object in an associated manner; The determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library includes: determining a target session vector according to the target session message; The object corresponding to the target conversation vector in the conversation vector library is determined as the object to be queried corresponding to the target conversation message.

9. The method according to claim 7, characterized in that The conversation message library also includes conversation keywords corresponding to the objects; and determining the object to be queried corresponding to the target conversation message in the pre-built conversation message library includes: determining a target conversation keyword of the target conversation message; Determining a first conversation keyword that matches the target conversation keyword in the conversation message library; The object corresponding to the first conversation keyword is determined as the to-be-queried object corresponding to the target conversation message.

10. The method according to any one of claims 1 to 9, characterized in that Outputting a target reply message corresponding to the target conversation message through a preset model according to the current state and the object to be queried includes: Determine the text of the current state and the text of the object to be queried as model input text; Inputting the model input text into the preset model to obtain a target reply message corresponding to the target conversation message; The target reply message is displayed.

11. The method according to claim 10, characterized in that The step of determining the text of the current state and the text of the object to be queried as model input text includes: Obtaining state switching guidance text related to the object to be queried; The state switching guide text, the text of the current state and the text of the object to be queried are determined as model input text.

12. A human-computer interaction device, characterized in that: The device is used to implement the steps of the human-computer interaction method as described in any one of claims 1 to 11.

13. A device for realizing human-computer interaction, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to execute the instructions to enable the device to perform the steps of the human-computer interaction method according to any one of claims 1 to 11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the human-computer interaction method according to any one of claims 1 to 11 are implemented.

15. A computer program product, characterized in that The invention comprises a computer program, which implements the steps of the human-computer interaction method according to any one of claims 1 to 11 when the computer program is executed by a processor.

Images