A human-computer conversation method, system and related device

Abstract

This application discloses a human-computer dialogue method, system, and related apparatus. The method includes: acquiring the current demand in the current round; using an intelligent analysis model to perform correlation analysis on the current demand and historical interaction content from at least some rounds to obtain a correlation analysis result; determining the content to be analyzed corresponding to the current round based on the correlation analysis result; wherein, when the correlation analysis result is relevant, the content to be analyzed includes the current demand and the historical interaction content related to the current demand; when the correlation analysis result is irrelevant, the content to be analyzed includes the current demand; and obtaining a target response matching the current demand based on the content to be analyzed. Through the above method, this application can improve the relevance between the target response and the user's needs.

Description

Technical Field

[0001] This application relates to the field of human-computer interaction technology, and in particular to a human-computer dialogue method, system and related device. Background Technology

[0002] With the continuous development of human-computer interaction technology, chatbots can provide corresponding target responses based on user requests. To improve the relevance between the generated response and historical interactions, chatbots typically combine all historical interactions to arrive at the target response for the current request. However, this method cannot filter all historical interactions to obtain those relevant to the current need, causing the chatbot to process redundant information. This increases computational costs and interferes with the arrival of the target response. Therefore, effectively filtering historical interactions and improving the relevance of the target response to the user's current needs has become an urgent problem to be solved. Summary of the Invention

[0003] The main technical problem addressed by this application is to provide a human-computer dialogue method system and related apparatus that can improve the alignment between the target response and the user's needs.

[0004] To address the aforementioned technical problems, this application provides a human-computer dialogue method, comprising: acquiring the current demand in the current round; performing a correlation analysis on the current demand and at least some historical interaction content from previous rounds using an intelligent analysis model to obtain a correlation analysis result; determining the content to be analyzed corresponding to the current round based on the correlation analysis result; wherein, when the correlation analysis result is relevant, the content to be analyzed includes the current demand and the historical interaction content related to the current demand; when the correlation analysis result is irrelevant, the content to be analyzed includes the current demand; and obtaining a target response matching the current demand based on the content to be analyzed.

[0005] To address the aforementioned technical problems, another technical solution adopted in this application is: providing a human-computer dialogue system, comprising: an analysis module, configured to acquire the current demand in the current round, and perform correlation analysis on the current demand and at least some historical interaction content from previous rounds using an intelligent analysis model to obtain a correlation analysis result; a processing module, configured to determine the content to be analyzed corresponding to the current round based on the correlation analysis result; wherein, when the correlation analysis result is relevant, the content to be analyzed includes the current demand and the historical interaction content related to the current demand, and when the correlation analysis result is irrelevant, the content to be analyzed includes the current demand; and a response module, configured to obtain a target response matching the current demand based on the content to be analyzed.

[0006] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a mobile terminal, including a display screen, a memory, and a processor. The display screen and the memory are respectively coupled to the processor. The display screen is used at least to display content to the user and to allow the user to input or select content. The memory stores program instructions, and the processor is used to execute the program instructions to implement the human-computer dialogue method mentioned in the above technical solution.

[0007] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a computer-readable storage medium storing program instructions thereon, wherein the program instructions, when executed by a processor, implement the human-computer dialogue method mentioned in the above technical solution.

[0008] The beneficial effects of this application are as follows: Unlike existing technologies, the human-computer dialogue method proposed in this application, after obtaining the user's current needs in the current round, performs a correlation analysis on the current needs and at least some historical interaction content from previous rounds to obtain the correlation analysis results. Based on these results, it is determined whether there is historical interaction content related to the current needs. If so, the relevant historical interaction content is combined to obtain a response that matches the target needs. This ensures a high degree of relevance between the target response and the current needs, while also avoiding interference from irrelevant historical interaction content in obtaining the target response, thus improving the accuracy of obtaining the target response. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0010] Figure 1 This is a flowchart illustrating one implementation method of the applicant's human-computer dialogue method;

[0011] Figure 2 This is a flowchart illustrating step S101 corresponding to one embodiment.

[0012] Figure 3 This is a schematic diagram of one implementation method of the applicant's human-computer dialogue method;

[0013] Figure 4 This is a flowchart illustrating step S103 corresponding to one implementation method;

[0014] Figure 5 This is a flowchart illustrating another implementation of the applicant's human-computer dialogue method;

[0015] Figure 6This is a schematic diagram of one embodiment of the applicant's human-computer dialogue system;

[0016] Figure 7 This is a schematic diagram of the structure of one embodiment of the mobile terminal of this application;

[0017] Figure 8 This is a schematic diagram of one embodiment of the computer-readable storage medium of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0019] In this paper, the terms "system" and "network" are often used interchangeably. The term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "many" in this paper means two or more.

[0020] The human-computer dialogue method proposed in this application relies on a chatbot, which can be an application on a smart terminal or at least a smart terminal with integrated text processing capabilities. The aforementioned smart terminal can be a mobile phone, tablet computer, or personal computer, among other devices.

[0021] Please see Figure 1 , Figure 1 This is a flowchart illustrating one embodiment of the applicant's human-computer interaction method, which includes:

[0022] S101: Obtain the current requirements of the current round, and use the intelligent analysis model to perform correlation analysis on the current requirements and the historical interaction content of at least some rounds to obtain the correlation analysis results.

[0023] In one embodiment, step S101 includes: obtaining the current demand for the current round input by the user. The current demand can be a voice command input by the user or a text command manually input by the user.

[0024] In one implementation scenario, in response to a user-inputted voice command as the current requirement, the voice command is acquired, speech recognition is performed on the voice command to obtain the corresponding converted text, and this converted text is used as the current requirement for the current round. Alternatively, in response to a user-inputted text command as the current requirement, the text corresponding to that text command is used as the current requirement for the current round.

[0025] It should be noted that in practical applications, during the dialogue with users, including the multiple rounds of needs raised by the users and the responses that match the users' needs, the latest needs raised by the users are taken as the current needs, and the needs before the current needs are taken as historical needs. Each historical need is matched with a corresponding historical response.

[0026] Further, please refer to Figure 2 , Figure 2 This is a flowchart illustrating step S101 in one embodiment. Step S101 specifically includes:

[0027] S201: Input the current demand and the historical interaction content of at least some of the previous rounds into the intelligent analysis model to obtain the current topic corresponding to the current demand, and at least one historical topic corresponding to the historical interaction content of at least some of the previous rounds.

[0028] In one embodiment, step S201 includes: using historical interaction content that does not include a masking identifier as available interaction content. The available interaction content includes at least the historical interaction content from the previous round.

[0029] Specifically, after obtaining the user's current input request, the historical interaction content from the previous round of the current request is taken as adjacent historical content; and, in response to the fact that at least part of the historical interaction content is related to the aforementioned adjacent historical content, the adjacent historical content and the historical interaction content related to the adjacent historical content are taken as available interaction content. The aforementioned available interaction content does not include masking indicators.

[0030] It should be noted that the aforementioned masking markers are used to mask corresponding historical interaction content during the subsequent generation of the target response, in order to avoid interference from historical interaction content with masking markers when generating the target response. Furthermore, the process of adding or removing masking markers from historical interaction content can be found in the corresponding implementation methods described below.

[0031] In a specific implementation scenario, please refer to Figure 3 , Figure 3This is a schematic diagram of one implementation method of the user-computer dialogue method. After obtaining the user's current requirement 40 in the current round, the first historical requirement 30 and the matching first historical response 31 from the previous round are used as available interactive content. Since the first historical requirement 30 is a requirement condition added by the user based on the second historical requirement 20 in the previous round, the first historical requirement 30 is related to the second historical requirement 20 in the previous round. In addition, in response to the fact that the second historical requirement 20 is not related to the third historical requirement 10 in the previous round, in this specific implementation scenario, the first historical requirement 30 and its matching first historical response 31, as well as the second historical requirement 20 and its matching second historical response 21, are used as available interactive content.

[0032] Furthermore, the current needs and available interactive content are input into the intelligent analysis model, and the intelligent analysis model is required to output the current topic corresponding to the current needs, and the historical topics corresponding to the available interactive content.

[0033] In one implementation scenario, the aforementioned intelligent analysis model is a large language model with superior language understanding and text generation capabilities. The current needs and available interactive content are provided to the large language model, and the large language model is required to perform semantic analysis on the current needs and available interactive content to output the current topic corresponding to the current needs and at least one historical topic corresponding to the available interactive content.

[0034] In a specific application scenario, the aforementioned large language model is a generative pre-trained Transformer model (GPT).

[0035] S202: Based on the current topic and at least one historical topic, determine the results of the correlation analysis.

[0036] In one embodiment, step S202 includes comparing the current topic corresponding to the current demand with at least one historical topic corresponding to historical interaction content to determine the correlation analysis result. The correlation analysis result includes whether the content is relevant or irrelevant.

[0037] In one implementation scenario, in response to historical interaction content, including available interaction content, the current topic corresponding to the current need is first compared with the historical topics corresponding to the available interaction content. If the current topic matches the historical topic corresponding to the available interaction content, it is determined that the current need is related to at least part of the historical interaction content, and no topic switch has occurred; that is, the correlation analysis result is determined to be related to the available interaction content. Subsequently, the target response corresponding to the current need needs to be obtained by combining the available interaction content, so that the target response has a high degree of consistency with the user's current need.

[0038] In another implementation scenario, in response to the distinction between the current topic and the historical topics corresponding to the available interactive content, it is determined that a topic switch has occurred. At least some rounds of historical interactive content, including the masking marker, are obtained as reference interactive content. All reference interactive content is input into the intelligent analysis model to obtain at least one reference topic corresponding to all reference interactive content.

[0039] Specifically, in response to a topic change due to a current demand, it is necessary to determine whether the current topic corresponding to the current demand is related to at least some of the historical interaction content preceding the available interactive content. The current timestamp corresponding to the current demand is obtained, and all historical interaction content, including those marked with a mask, within a preset time period prior to the current timestamp is used as reference interaction content. An intelligent analysis model is then used to obtain at least one reference topic corresponding to the reference interaction content. For example, if the preset time period is 2 hours, then all historical interaction content, including those marked with a mask, within two hours of the current timestamp is used as reference interaction content. The preset time period can also be other values, and can be set according to actual needs; this application does not limit this.

[0040] Optionally, the historical interaction content including the masking markers from the first preset number of rounds prior to the current round can be used as the reference interaction content. For example, if the first preset number is ten, then the historical interaction content including the masking marker pairs from the ten rounds prior to the current round corresponding to the current requirement can be used as the reference interaction content. Alternatively, the historical interaction content including the masking markers from the second preset number of rounds prior to the current round and closest to the current round can be used as the reference interaction content. For example, if the second preset number is five, then the five historical interaction contents including the masking markers from the current round can be obtained sequentially in order of proximity to the current round as the reference interaction content. It should be noted that the specific values ​​of the first and second preset numbers can be set according to the actual situation, and this application does not limit them.

[0041] In a specific implementation scenario, please refer to [the relevant documentation]. Figure 3 Since the current topic corresponding to current demand 40 is the first topic, the available interactive content, including the first historical demand 30, the first historical reply 31, the second historical demand 20, and the second historical reply 21, corresponds to a second topic different from the first topic, meaning the current demand has undergone a topic switch. At this point, the third historical demand 10 and the matching third historical reply 11 from previous rounds of available interactive content are used as reference interactive content, and an intelligent analysis model is used to obtain the reference topic corresponding to the reference interactive content.

[0042] Furthermore, in response to the current topic being consistent with any reference topic, it is determined that the current demand is related to at least some of the historical interaction content, that is, the determined correlation analysis result is related to the reference interaction content of each round consistent with the current topic; at the same time, the masking labels of the reference interaction content of each round consistent with the current topic are removed, and masking labels are added to the available interaction content.

[0043] For details, please continue reading Figure 3 In response to the fact that the reference topic corresponding to the third historical request 10 and the third historical reply 11 in the reference interaction content is the first topic, that is, consistent with the current topic corresponding to the current request 40, it is determined that the current request 40 is related to the third historical request 10 and the third historical reply 11, and the blocking mark of the third historical request 10 and the third historical reply 11 is removed; and, the available interaction content is marked with a blocking mark, that is, the first historical request 30, the first historical reply 31, the second historical request 20 and the second historical reply 21 are marked with a blocking mark.

[0044] Alternatively, in response to the distinction between the current topic and all reference topics, it is determined that the current requirement is unrelated to historical interaction content, that is, the correlation analysis result is determined to be unrelated to all interaction content, and a masking label is added to the available interaction content.

[0045] The above solution allows the generation of the target response for the current need based solely on historical interaction content that is not marked with a blocking indicator and the current need. This avoids combining all historical interaction content, saving computational costs and preventing irrelevant information from interfering with the acquisition of the target response, thus improving the accuracy of obtaining the target response.

[0046] S102: Based on the correlation analysis results, determine the content to be analyzed for the current round. When the correlation analysis result is relevant, the content to be analyzed includes the current requirement and related historical interactions; when the correlation analysis result is irrelevant, the content to be analyzed includes the current requirement.

[0047] In one implementation scenario, in response to the correlation analysis result being relevant, i.e. the current requirement is related to the historical interaction content of at least some rounds, the current requirement and the historical interaction content related to the current requirement are taken as the content to be analyzed for the current round.

[0048] In another implementation scenario, if the correlation analysis result is irrelevant, meaning that the current requirement is irrelevant to the historical interaction content of all rounds, then only the current requirement will be considered as the content to be analyzed.

[0049] S103: Based on the content to be analyzed, obtain the target response that matches the current requirements.

[0050] In one implementation, please refer to Figure 4 , Figure 4 This is a flowchart illustrating step S103 in one embodiment. Step S103 specifically includes:

[0051] S301: Input the content to be analyzed into the intelligent analysis model to obtain the first candidate response output by the intelligent analysis model; and, based on the content to be analyzed, predict the response content to obtain the second candidate response.

[0052] In one implementation scenario, in response to the correlation analysis result being relevant, the current requirement and related historical interaction content are input into the intelligent analysis model to obtain the first candidate response output by the intelligent analysis model. Based on the current requirement and related historical interaction content, the response content is estimated to obtain the second candidate response.

[0053] In a specific implementation scenario, the chatbot includes a response prediction model. After determining the content to be analyzed, the response prediction model performs semantic analysis on the content and, in conjunction with relevant data, generates a second candidate response that matches the current need. This relevant data can be pre-stored user-related information or information related to the current need obtained through the network. The response prediction model is trained using multiple training datasets, including training needs labeled with response tags. Furthermore, the specific structure of this response prediction model can refer to existing semantic analysis model structures, which will not be elaborated upon in this application.

[0054] In another implementation scenario, in response to the irrelevance analysis result, the current requirement is input into the intelligent analysis model, resulting in a first candidate response output by the model. Based on the current requirement, the response content is estimated to obtain a second candidate response. Unlike the previous implementation scenario, since the correlation analysis result is irrelevant, it is not necessary to consider historical interaction content when obtaining the target response corresponding to the current requirement. The generation process of the first and second candidate responses in this scenario can refer to the implementation scenario described above.

[0055] S302: Based on the first candidate response and the second candidate response, obtain the target response.

[0056] In one embodiment, step S302 includes: inputting the current topic corresponding to the current need, the first candidate response, and the second candidate response into the intelligent analysis model, so that the intelligent analysis model selects the response with a higher degree of fit with the current need from the first candidate response and the second candidate response as the target response. After obtaining the target response output by the intelligent analysis model, the target response is displayed to the user on the display interface.

[0057] In one implementation scenario, the intelligent analysis model determines the relevance scores of a first candidate response and a second candidate response to the current topic. If the relevance score of the first candidate response is greater than that of the second candidate response, the intelligent analysis model outputs the first candidate response as the target response. Alternatively, if the relevance score of the first candidate response is less than that of the second candidate response, the intelligent analysis model outputs the second candidate response as the target response. Or, if the relevance scores of the first and second candidate responses are the same, either the first or second candidate response is arbitrarily selected as the target response.

[0058] In another implementation scenario, after the current topic, the first candidate response, and the second candidate response are input into the intelligent analysis model, the intelligent analysis model determines the degree of fit between the first candidate response and the second candidate response and the current requirement, and takes the response with the highest degree of fit as the target response.

[0059] In another implementation scenario, the intelligent analysis model outperforms chatbots in language understanding and text generation, while chatbots store user-related information during user interactions. After inputting the current topic, first candidate response, and second candidate response into the intelligent analysis model, the model compares the current topic with historical topics. If the current topic matches any historical topic, meaning the current need is related to at least some historical interactions, the intelligent analysis model uses its first candidate response as the target response to ensure a higher degree of relevance to the user's current need. If the current topic differs from any historical topic, meaning the current need is unrelated to all historical interactions, the intelligent analysis model uses the second candidate response as the target response to ensure a higher degree of relevance to the user's information.

[0060] The human-computer dialogue method proposed in this application, after obtaining the user's current needs in the current round, performs a correlation analysis on the current needs and at least some historical interaction content from previous rounds to obtain the correlation analysis results. Based on these results, it determines whether there is historical interaction content related to the current needs. If so, it combines the relevant historical interaction content to obtain a target response, ensuring a high degree of relevance between the target response and the current needs while avoiding interference from irrelevant historical interaction content in obtaining the target response, thus improving the accuracy of obtaining the target response.

[0061] In another implementation, please refer to Figure 5 , Figure 5 This is a flowchart illustrating another embodiment of the applicant's human-computer interaction method. This embodiment specifically includes:

[0062] S401: Obtain the current requirements of the current round, and use the intelligent analysis model to perform correlation analysis on the current requirements and the historical interaction content of at least some rounds to obtain the correlation analysis results.

[0063] In one embodiment, the implementation process of step S401 can be referred to Figure 1 Step S101 will not be described in detail here.

[0064] S402: Obtain the number of interactions corresponding to the historical interaction content of all rounds related to the current topic, and input the number of interactions into the intelligent analysis model.

[0065] In one implementation, the number of interactions indicates the number of times the topic has switched between all rounds of historical interactions related to the current topic and the current requirement. By obtaining this number of interactions to represent the relevance between all rounds of historical interactions related to the current topic and the current requirement, the accuracy of decision instructions output by the subsequent intelligent analysis model can be improved.

[0066] In one specific implementation, please refer to [the document / document]. Figure 3 Since the third historical request 10 and the third historical reply 11 are historical interaction content related to the current first topic, and the topic has changed twice between the historical interaction content and the current request 40, the number of interactions corresponding to the third historical request 10 and the third historical reply 11 is 2.

[0067] Alternatively, in another implementation, the number of interactions can be the difference between the current round of the current requirement and the round corresponding to the historical interaction content. Specifically, since the historical interaction content for all rounds related to the current topic includes multiple adjacent historical interaction contents, the minimum difference between the current round and the round corresponding to the historical interaction content is taken as the number of interactions for all rounds related to the current topic.

[0068] S403: Obtain the decision instruction output by the intelligent analysis model based on the current topic and the number of interactions. The decision instruction indicates whether to utilize the intelligent analysis model to obtain a target response that matches the current needs.

[0069] In one embodiment, the implementation process of step S403 includes: after inputting the number of interactions and the current topic into the intelligent analysis model, requesting the intelligent analysis model to output a decision instruction based on the current topic and the number of interactions.

[0070] In another embodiment, the decision instruction in step S403 can also be generated by the chatbot based on the current topic and the number of interactions. The process includes: If the number of interactions is less than a threshold, it indicates that the corresponding historical interaction content has fewer topic switches than the current need, and the correlation between the corresponding historical interaction content and the current need is high. In this case, a decision instruction is generated to obtain the target response using an intelligent analysis model, which helps to subsequently obtain the target response using an intelligent analysis model with better language understanding and text generation capabilities, thereby improving the accuracy of the target response. If the number of interactions is greater than or equal to the threshold, it indicates that the corresponding historical interaction content has more topic switches than the current need, and the correlation between the corresponding historical interaction content and the current need is low. In this case, a decision instruction is generated to generate the target response using the chatbot's own processor, which helps to subsequently obtain the target response directly using the chatbot's own processor, thereby improving the efficiency of obtaining the target response.

[0071] S404: Based on the correlation analysis results, determine the content to be analyzed for the current round.

[0072] In one embodiment, the implementation process of step S404 includes: in response to the correlation analysis result being relevant, taking the current requirement and the historical interaction content related to the current requirement as the content to be analyzed for the current round. Alternatively, in response to the correlation analysis result being irrelevant, that is, the current requirement is irrelevant to the historical interaction content of all rounds, then only the current requirement will be taken as the content to be analyzed.

[0073] S405: Based on the content to be analyzed, obtain the target response that matches the current requirements.

[0074] In one embodiment, step S405 includes obtaining a target response matching the current need based on a decision instruction and the content to be analyzed. The decision instruction includes either a first identifier or a second identifier. When the decision instruction includes the first identifier, an intelligent analysis model is used to obtain a target response matching the current need; when the decision instruction includes the second identifier, the chatbot's own processor obtains a target response matching the current need based on the content to be analyzed.

[0075] In one implementation scenario, in response to a correlation analysis result indicating relevance and a decision instruction including a first identifier, the current need and related historical interactions are input into the intelligent analysis model. This model leverages its superior language understanding and text generation capabilities to output a target response. The specific process of outputting the target response using the intelligent analysis model can be found in the corresponding implementation described above and will not be elaborated upon here.

[0076] In another implementation scenario, in response to the correlation analysis result being relevant and the decision instruction including a second identifier, the response content is estimated based on the current need and historical interaction content related to the current need, to obtain the target response.

[0077] The target response is obtained by using the response prediction model contained in the chatbot based on the current need and the historical interaction content related to the current need. The specific process can be referred to the corresponding implementation method described above, and will not be elaborated here.

[0078] In another implementation scenario, in response to the correlation analysis result being irrelevant and the decision instruction including the first identifier, the current requirement is input into the intelligent analysis model to obtain the target response output by the intelligent analysis model. The specific process can be referred to the above implementation scenario.

[0079] In another implementation scenario, in response to the correlation analysis result being irrelevant and the decision instruction including a second identifier, the response content is estimated based on the current needs to obtain the target response. The specific process can be referred to the above implementation scenario.

[0080] In another embodiment, the step of obtaining the current requirements of the current round in step S101 may further include: obtaining audio information in the current scene, and obtaining the transcribed text corresponding to the current scene based on the audio information.

[0081] Specifically, the audio information responding to the current scenario includes audio of conversations between the user and other people, as well as audio of the user's requests to the chatbot. Based on speech recognition technology, the audio information in the current scenario is identified to obtain the corresponding transcribed text. This transcribed text contains both the user's requests to the chatbot and the user's conversations with other people in the current scenario.

[0082] Furthermore, the transcribed text is input into the intelligent analysis model to obtain the current requirements for the current round, as output by the intelligent analysis model.

[0083] Specifically, an intelligent analysis model is used to perform semantic analysis on the transcribed text to identify the user's requests to the chatbot, which are then used as the current request for the current round. This method can automatically identify user requests from audio in complex scenarios without requiring users to provide trigger words before each request, thus improving the efficiency of user request identification.

[0084] In another embodiment, the implementation process of step S201 may further include: inputting the current demand and the historical interaction content of at least some previous rounds into the intelligent analysis model to obtain the current topic corresponding to the current demand, and at least one historical topic corresponding to the input historical interaction content.

[0085] Furthermore, if the current topic matches any historical topic, then the current need is determined to be related to at least some of the historical interaction content, and the correlation analysis result is "related". Alternatively, if the current topic differs from all historical topics in terms of relevance, then a topic switch is determined to have occurred, and the correlation analysis result is "irrelevant".

[0086] Furthermore, based on the correlation analysis results, the content to be analyzed corresponding to the current round is determined, and based on the content to be analyzed, a target response matching the current needs is obtained. The specific processes for determining the content to be analyzed and obtaining the target response can be referred to the corresponding implementation methods described above, and will not be elaborated upon here.

[0087] Please see Figure 6 , Figure 6 This is a schematic diagram of one embodiment of the human-computer dialogue system proposed by the applicant. The human-computer dialogue system includes an analysis module 50, a processing module 60, and a response module 70.

[0088] Specifically, the analysis module 50 is used to obtain the current requirements of the current round, and to use an intelligent analysis model to perform correlation analysis on the current requirements and the historical interaction content of at least some rounds to obtain the correlation analysis results.

[0089] The processing module 60 is used to determine the content to be analyzed for the current round based on the correlation analysis results. Specifically, when the correlation analysis result is relevant, the content to be analyzed includes the current requirement and historical interaction content related to the current requirement; when the correlation analysis result is irrelevant, the content to be analyzed includes the current requirement.

[0090] The response module 70 is used to obtain a target response that matches the current requirements based on the content to be analyzed.

[0091] In one embodiment, the analysis module 50 uses an intelligent analysis model to perform a correlation analysis on the current demand and the historical interaction content of at least some rounds to obtain the correlation analysis results, including: inputting the current demand and the historical interaction content of at least some rounds before it into the intelligent analysis model to obtain the current topic corresponding to the current demand and at least one historical topic corresponding to the historical interaction content of at least some rounds; and determining the correlation analysis results based on the current topic and at least one historical topic.

[0092] In one implementation scenario, please refer to [the relevant documentation / reference]. Figure 6In the human-computer dialogue system proposed in this application, the analysis module 50 further includes a topic acquisition submodule 51, which is used to take historical interaction content without masking markers as available interaction content; wherein, the available interaction content includes at least the historical interaction content of the previous round; the current demand and available interaction content are input into the intelligent analysis model to obtain the current topic corresponding to the current demand and the historical topic corresponding to the available interaction content.

[0093] In another implementation scenario, please refer to [link / reference]. Figure 6 In the human-computer dialogue system proposed in this application, the analysis module 50 further includes a transcription submodule 52, which is used to obtain audio information in the current scene, obtain the corresponding transcribed text in the current scene based on the audio information, and input the transcribed text into the intelligent analysis model to obtain the current requirements of the current round output by the intelligent analysis model.

[0094] In one implementation scenario, the analysis module 50 determines the relevance analysis results based on the current topic and at least one historical topic, including: responding to the current topic being consistent with the historical topic corresponding to the available interactive content, determining that the current requirement is related to at least some of the historical interactive content and that no topic switch has occurred; responding to the current topic being distinct from the historical topics corresponding to the available interactive content, determining that a topic switch has occurred, obtaining at least some rounds of historical interactive content including masking tags as reference interactive content, inputting all reference interactive content into the intelligent analysis model to obtain at least one reference topic corresponding to all reference interactive content; responding to the current topic being consistent with any reference topic, determining that the current requirement is related to at least some of the historical interactive content, removing the masking tags of the reference interactive content in each round consistent with the current topic, and adding masking tags to the available interactive content; responding to the current topic being distinct from all reference topics, determining that the current requirement is unrelated to the historical interactive content, and adding masking tags to the available interactive content.

[0095] In a specific implementation scenario, the analysis module 50 obtains historical interaction content including masking markers from at least several rounds as reference interaction content, including: obtaining the current timestamp corresponding to the current requirement, and using all historical interaction content including masking markers within a preset time length before the current timestamp as reference interaction content; or, using historical interaction content including masking markers from a preset number of rounds before the current round as reference interaction content.

[0096] In another embodiment, please refer to [link / reference needed]. Figure 6The human-computer dialogue system proposed in this application also includes a decision instruction generation submodule 71 coupled to the response module 70. After obtaining the relevance analysis results, the decision instruction generation submodule 71 is used to obtain the number of interactions corresponding to the historical interaction content of all rounds related to the current topic, and input the number of interactions into the intelligent analysis model; obtain the decision instruction output by the intelligent analysis model based on the current topic and the number of interactions; wherein, the decision instruction is used to indicate whether to use the intelligent analysis model to obtain a target response that matches the current needs.

[0097] In one implementation scenario, the decision instruction includes a first identifier or a second identifier. The response module 70 obtains a target response matching the current need based on the content to be analyzed, including: responding to a relevance analysis result that is relevant and the decision instruction includes the first identifier, inputting the current need and historical interaction content related to the current need into the intelligent analysis model to obtain the target response output by the intelligent analysis model; responding to a relevance analysis result that is relevant and the decision instruction includes the second identifier, estimating the response content based on the current need and historical interaction content related to the current need to obtain the target response; responding to a non-relevance analysis result that is irrelevant and the decision instruction includes the first identifier, inputting the current need into the intelligent analysis model to obtain the target response output by the intelligent analysis model; and responding to a non-relevance analysis result that is irrelevant and the decision instruction includes the second identifier, estimating the response content based on the current need to obtain the target response.

[0098] In another embodiment, the response module 70 obtains a target response matching the current need based on the content to be analyzed, including: in response to a relevance analysis result that is relevant, inputting the current need and historical interaction content related to the current need into the intelligent analysis model to obtain a first candidate response output by the intelligent analysis model; estimating the response content based on the current need and historical interaction content related to the current need to obtain a second candidate response; in response to a relevance analysis result that is irrelevant, inputting the current need into the intelligent analysis model to obtain a first candidate response output by the intelligent analysis model; estimating the response content based on the current need to obtain a second candidate response; and obtaining the target response based on the first candidate response and the second candidate response.

[0099] In one implementation scenario, the response module 70 obtains the target response based on the first candidate response and the second candidate response, including: inputting the current topic, the first candidate response, and the second candidate response into the intelligent analysis model to obtain the target response output by the intelligent analysis model.

[0100] Please see Figure 7 , Figure 7This is a schematic diagram of the structure of a mobile terminal according to an embodiment of this application. The mobile terminal includes a display screen 80, a memory 90, and a processor 100. The display screen 80 and the memory 90 are respectively coupled to the processor 100. The display screen 80 is used at least to display content to the user and to allow the user to input or select content. The memory 90 stores program instructions, and the processor 100 is used to execute the program instructions to implement the human-computer dialogue method mentioned in any of the above embodiments. Specifically, electronic devices include, but are not limited to, desktop computers, laptops, tablets, servers, etc., and are not limited here. In addition, the processor 100 may also be called a CPU (Center Processing Unit). The processor 100 may be an integrated circuit chip with signal processing capabilities. The processor 100 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor may be a microprocessor or any conventional processor. In addition, the processor 100 can be implemented by integrated circuit chips.

[0101] Please see Figure 8 , Figure 8 This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present application. The computer-readable storage medium 110 stores program instructions 120 that can be executed by a processor. When the program instructions 120 are executed by the processor, they implement the human-computer dialogue method mentioned in any of the above embodiments.

[0102] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0103] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0104] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0105] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0106] The above are merely embodiments of this application and do not limit the scope of this patent application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A human-computer dialogue method, characterized in that, include: Obtain the current demand for the current round, and use an intelligent analysis model to perform a correlation analysis on the current demand and the historical interaction content of at least some rounds to obtain the correlation analysis result; the correlation analysis result is obtained based on the following steps: input the current demand and the historical interaction content of at least some previous rounds into the intelligent analysis model to obtain the current topic corresponding to the current demand, and at least one historical topic corresponding to the historical interaction content of at least some rounds; determine the correlation analysis result based on the current topic and at least one historical topic; Based on the correlation analysis results, the content to be analyzed corresponding to the current round is determined; wherein, when the correlation analysis result is relevant, the content to be analyzed includes the current requirement and the historical interaction content related to the current requirement; when the correlation analysis result is irrelevant, the content to be analyzed includes the current requirement. Based on the content to be analyzed, a target response matching the current requirement is obtained; The process of obtaining the current demand for the current round, performing correlation analysis on the current demand and at least some historical interaction content from previous rounds using an intelligent analysis model, and obtaining the correlation analysis results includes: Obtain the number of interactions corresponding to the historical interaction content of all rounds related to the current topic, and input the number of interactions into the intelligent analysis model; Obtain the decision instruction output by the intelligent analysis model based on the current topic and the number of interactions; wherein, the decision instruction is used to indicate whether to use the intelligent analysis model to obtain a target response that matches the current need.

2. The method according to claim 1, characterized in that, The step of inputting the current demand and at least some of the previous rounds of historical interaction content into the intelligent analysis model to obtain the current topic corresponding to the current demand and at least one historical topic corresponding to at least some of the previous rounds of historical interaction content includes: The historical interaction content that does not include the masking identifier is regarded as the available interaction content; wherein, the available interaction content includes at least the historical interaction content of the previous round; The current demand and the available interactive content are input into the intelligent analysis model to obtain the current topic corresponding to the current demand and the historical topic corresponding to the available interactive content.

3. The method according to claim 2, characterized in that, The determination of the correlation analysis results based on the current topic and at least one of the historical topics includes: In response to the fact that the current topic is consistent with the historical topic corresponding to the available interactive content, it is determined that the current requirement is related to at least part of the historical interactive content and no topic switch has occurred; In response to the difference between the current topic and the historical topic corresponding to the available interactive content, it is determined that a topic switch has occurred, and the historical interactive content including the blocking identifier in at least some rounds is obtained as reference interactive content. All the reference interactive content is input into the intelligent analysis model to obtain at least one reference topic corresponding to all the reference interactive content. In response to the current topic being consistent with any of the reference topics, it is determined that the current requirement is related to at least part of the historical interaction content, the masking identifier of the reference interaction content in each round that is consistent with the current topic is removed, and the masking identifier is added to the available interaction content; Since the current topic is distinct from all the reference topics, it is determined that the current requirement is unrelated to the historical interaction content, and the masking identifier is added to the available interaction content.

4. The method according to claim 3, characterized in that, The step of obtaining the historical interaction content, including at least a portion of the rounds with the masking identifier, as reference interaction content includes: Obtain the current timestamp corresponding to the current requirement, and use all historical interaction content including the blocking identifier within a preset time length before the current timestamp as reference interaction content; or, The historical interaction content, including the masking identifier, from a preset number of rounds prior to the current round is used as the reference interaction content.

5. The method according to claim 1, characterized in that, The decision instruction includes a first identifier or a second identifier, and the step of obtaining a target response matching the current requirement based on the content to be analyzed includes: In response to the correlation analysis result being relevant, and the decision instruction including the first identifier, the current need and the historical interaction content related to the current need are input into the intelligent analysis model to obtain the target response output by the intelligent analysis model; In response to the correlation analysis result being relevant, and the decision instruction including the second identifier, the response content is estimated based on the current need and the historical interaction content related to the current need, to obtain the target response; In response to the correlation analysis result being irrelevant, and the decision instruction including the first identifier, the current requirement is input into the intelligent analysis model to obtain the target response output by the intelligent analysis model; In response to the correlation analysis result being irrelevant, and the decision instruction including the second identifier, the response content is estimated based on the current requirement to obtain the target response.

6. The method according to claim 1, characterized in that, The process of obtaining a target response that matches the current requirement based on the content to be analyzed includes: In response to the correlation analysis result being relevant, the current need and the historical interaction content related to the current need are input into the intelligent analysis model to obtain the first candidate response output by the intelligent analysis model. Based on the current need and the historical interaction content related to the current need, the response content is estimated to obtain the second candidate response. In response to the correlation analysis result being irrelevant, the current requirement is input into the intelligent analysis model to obtain the first candidate response output by the intelligent analysis model. Based on the current requirement, the response content is estimated to obtain the second candidate response. The target response is obtained based on the first candidate response and the second candidate response.

7. The method according to claim 6, characterized in that, The step of obtaining the target response based on the first candidate response and the second candidate response includes: The current topic, the first candidate response, and the second candidate response are input into the intelligent analysis model to obtain the target response output by the intelligent analysis model.

8. The method according to claim 1, characterized in that, The process of obtaining the current requirements for the current round includes: Obtain audio information for the current scene, and based on the audio information, obtain the transcribed text corresponding to the current scene; The transcribed text is input into the intelligent analysis model to obtain the current requirements for the current round, as output by the intelligent analysis model.

9. A human-computer dialogue system, characterized in that, include: An analysis module is used to obtain the current demand in the current round, and to perform a correlation analysis on the current demand and the historical interaction content of at least some rounds using an intelligent analysis model to obtain the correlation analysis result. The correlation analysis result is obtained based on the following steps: inputting the current demand and the historical interaction content of at least some rounds before it into the intelligent analysis model to obtain the current topic corresponding to the current demand, and at least one historical topic corresponding to the historical interaction content of at least some rounds; and determining the correlation analysis result based on the current topic and at least one historical topic. The processing module is used to determine the content to be analyzed corresponding to the current round based on the correlation analysis results; wherein, when the correlation analysis results are relevant, the content to be analyzed includes the current requirement and the historical interaction content related to the current requirement; when the correlation analysis results are irrelevant, the content to be analyzed includes the current requirement. The response module is used to obtain a target response that matches the current requirement based on the content to be analyzed. The process of obtaining the current demand for the current round, performing correlation analysis on the current demand and at least some historical interaction content from previous rounds using an intelligent analysis model, and obtaining the correlation analysis results includes: Obtain the number of interactions corresponding to the historical interaction content of all rounds related to the current topic, and input the number of interactions into the intelligent analysis model; Obtain the decision instruction output by the intelligent analysis model based on the current topic and the number of interactions; wherein, the decision instruction is used to indicate whether to use the intelligent analysis model to obtain a target response that matches the current need.

10. A mobile terminal, characterized in that, include: The system includes a display screen, a memory, and a processor, wherein the display screen and the memory are respectively coupled to the processor, the display screen is used at least to display content to a user and to allow the user to input or select content, the memory stores program instructions, and the processor is used to execute the program instructions to implement the human-computer dialogue method as described in any one of claims 1 to 8.

11. A computer-readable storage medium having program instructions stored thereon, characterized in that, When the program instructions are executed by the processor, they implement the human-computer dialogue method as described in any one of claims 1 to 8.

Images