Conversation information optimization method, electronic device and program product

Abstract

The embodiment of the invention discloses a dialogue information optimization method, electronic equipment and a program product, historical dialogue information corresponding to a current task is acquired in response to the end of a round of dialogue in the process of processing the current task by an intelligent agent, and the historical dialogue information comprises at least one round of dialogue information generated by the intelligent agent in at least one latest round; each round of dialogue information comprises task description information input to the intelligent agent; performing content analysis on the historical dialogue information to obtain a content score of the historical dialogue information; and in response to the content score indicating that similar dialogue information and / or invalid dialogue information exist in the historical dialogue information, updating previous task description information input to the intelligent agent in the last round of dialogue to obtain updated task description information, so that the intelligent agent performs the next round of dialogue based on the updated task description information. The quality of the content finally generated by the intelligent agent is improved, the convergence speed of the intelligent agent is improved, and the task processing efficiency is improved.

Description

Technical Field

[0001] The present disclosure relates to artificial intelligence technology, natural language processing technology, and in particular to a method for optimizing dialogue information, an electronic device, and a program product. Background Art

[0002] In an agent system based on a large language model (LLM), the agent can iteratively process tasks and obtain task processing results through multiple rounds of iterations to improve the quality of task processing.

[0003] In the process of iterative task processing by the intelligent agent, the related technology usually sets a fixed number of iterations to end the iteration process of the intelligent agent to avoid an infinite loop of the intelligent agent. However, during multiple iterations, the task description information (prompt) included in the dialogue information generated by the intelligent agent multiple times is completely consistent, making it impossible for the intelligent agent's multiple iterations to generate a better solution, and the dialogue cannot converge, which affects the quality of the intelligent agent's output results. The intelligent agent falls into unnecessary loops, resulting in a waste of computing resources and increased time and economic costs. Summary of the invention

[0004] The embodiments of the present disclosure provide a method for optimizing dialogue information, an electronic device, a storage medium, and a program product to solve the above technical problems to a certain extent.

[0005] One aspect of an embodiment of the present disclosure provides a method for optimizing dialogue information, the method comprising:

[0006] In the process of the agent processing the task, in response to the end of a round of dialogue, historical dialogue information corresponding to the task is obtained, wherein the historical dialogue information includes at least one round of dialogue information generated by the agent in at least one recent round, and each round of dialogue information includes task description information input to the agent;

[0007] Performing content analysis on the historical conversation information to obtain a content score of the historical conversation information, wherein the content score is used to characterize at least one of similarity and effectiveness of the historical conversation information;

[0008] In response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, the previous task description information input to the agent in the most recent round of dialogue is updated to obtain updated task description information, so that the agent conducts the next round of dialogue based on the updated task description information.

[0009] In an exemplary embodiment, performing content analysis on the historical conversation information to obtain a content score for the historical conversation information includes:

[0010] Performing a similarity test on the historical dialogue information to obtain a similarity score for the historical dialogue information, wherein the similarity score is used to represent the degree of similarity between dialogue information generated in different rounds;

[0011] and / or,

[0012] The historical dialogue information is tested for validity to obtain a validity score for the historical dialogue information, where the validity score is used to characterize the degree of relevance between the dialogue information and the task objective to be processed by the agent.

[0013] In an exemplary embodiment, performing similarity detection on the historical conversation information to obtain a similarity score of the historical conversation information includes:

[0014] Obtaining a task tag included in each round of dialogue information in the historical dialogue information, wherein the task tag is used to indicate a processing task corresponding to the dialogue information; and determining a similarity score of the historical dialogue information based on the number of repetitions of the task tag;

[0015] and / or,

[0016] A similarity test is performed on the information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

[0017] In an exemplary embodiment, after obtaining the task label included in each round of dialogue information in the historical dialogue information, the method further includes:

[0018] In response to the repetition number of the task tag reaching a number threshold, similarity detection is performed on information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

[0019] In an exemplary embodiment, in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, updating the last task description information input to the agent in the most recent round of dialogue includes:

[0020] In response to the similarity score being higher than a similarity threshold, determining that the historical dialogue information contains the similar dialogue information, and updating the last task description information input to the agent in the most recent round of dialogue;

[0021] and / or,

[0022] In response to the validity score being lower than a validity threshold, it is determined that the historical dialogue information contains the invalid dialogue information, and the last task description information input to the agent in the most recent round of dialogue is updated.

[0023] In an exemplary embodiment, the updating of the last task description information input to the agent in the most recent round of dialogue includes:

[0024] In response to the similarity score being higher than a similarity threshold, adding first guidance information to the previous task description information, wherein the first guidance information is used to indicate that the agent dialogue information is repeated;

[0025] In response to the effectiveness score being lower than the effectiveness threshold, second guidance information is added to the previous task description information, wherein the second guidance information is used to indicate that the agent task description information includes information that is irrelevant to the task goal.

[0026] In an exemplary embodiment, in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, the last task description information input to the agent in the most recent round of dialogue is updated:

[0027] In response to the content score indicating that similar conversation information and / or invalid conversation information exists in the historical conversation information, determining a similarity level and / or an invalidity level of the historical conversation information based on the content score;

[0028] Determining update content for updating the previous task description information based on the similarity level and / or the invalidity level, wherein the similarity level and / or the invalidity level is positively correlated with at least one of the amount of information content contained in the update content and the intensity of intervention of the update content on the task processed by the agent;

[0029] Based on the updated content, the last task description information input to the agent in the most recent round of dialogue is updated.

[0030] In an exemplary embodiment, after determining the similarity level and / or invalidity level of the historical conversation information based on the content score, the method further includes:

[0031] In response to the similarity level belonging to a preset severe similarity level, sending a dialogue termination instruction to the agent so that the agent stops iterating;

[0032] and / or,

[0033] In response to the invalidity level belonging to the preset serious invalidity level, a dialogue termination instruction is sent to the agent to stop the iteration of the agent.

[0034] In an exemplary embodiment, the updating of the last task description information input to the agent in the most recent round of dialogue includes:

[0035] In the previous task description information, at least one of guidance information for guiding the intelligent agent to modify the task description information, constraint conditions on the previous task description information, and newly added task description information is added.

[0036] Another aspect of the present embodiment provides a system for optimizing dialogue information, the system comprising an adjustment module and at least one intelligent agent;

[0037] The adjustment module is used to obtain historical dialogue information corresponding to the task in response to the end of a round of dialogue during the process of the at least one agent processing the task, wherein the historical dialogue information includes at least one round of dialogue information generated by the at least one agent in at least one recent round, and each round of dialogue information includes task description information input to the at least one agent; perform content analysis on the historical dialogue information to obtain a content score of the historical dialogue information, wherein the content score is used to characterize at least one of the similarity and validity of the historical dialogue information; in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, update the previous task description information input to the at least one agent in the most recent round of dialogue to obtain updated task description information, and send the updated task description information to the at least one agent;

[0038] The at least one intelligent agent is used to receive the updated task description information and conduct a next round of dialogue based on the updated task description information.

[0039] Another aspect of this embodiment provides an electronic device, including:

[0040] Memory for storing computer programs;

[0041] The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the method for optimizing the dialogue information described in any one of the above embodiments is implemented.

[0042] Another aspect of this embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for optimizing the dialogue information described in any of the above embodiments is implemented.

[0043] Another aspect of this embodiment provides a computer program product, including a computer program, which, when executed by a processor, implements the method for optimizing the conversation information described in any of the above embodiments.

[0044] In the disclosed embodiment, during the process of the intelligent agent processing this task, the dialogue information iteratively generated by the intelligent agent is monitored. After a round of dialogue ends, at least one round of historical dialogue information generated by the intelligent agent in at least one recent round of processing this task is obtained. The historical dialogue information can be content detected to obtain at least one of the similarity and validity of the historical dialogue information. When similar dialogue information and / or invalid dialogue information are detected in the historical dialogue information, the previous task description information input to the intelligent agent in the most recent round of dialogue can be updated to provide different task description information for the intelligent agent to intervene in the iterative process of the intelligent agent, which helps to improve the quality of the content finally generated by the intelligent agent. Compared with the fixed round number limitation method in the related art, the disclosed embodiment can dynamically adjust the task description information input to the intelligent agent by monitoring and analyzing dialogue information of different rounds, which can guide the dialogue of the intelligent agent to develop in the direction of convergence, avoid the intelligent agent from falling into an invalid cycle, help to improve the convergence speed of the intelligent agent, improve the task processing efficiency, and reduce the waste of computing resources.

[0045] In addition, through the mechanism of dynamically monitoring dialogue information and adaptively adjusting task description information, the quality of the content generated by the intelligent agent and the dialogue convergence speed can be balanced. While ensuring the high quality of the generated content, the dialogue convergence speed of the intelligent agent can be improved, which helps to improve the intelligent agent's ability to handle complex tasks, achieve a balance between the quality of generated content and dialogue convergence, and help improve the stability of the system.

[0046] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0048] The present disclosure may be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0049] Figure 1 A structural diagram of a system for optimizing dialogue information provided by an exemplary embodiment of the present disclosure;

[0050] Figure 2 A flowchart of a method for optimizing dialogue information provided by an exemplary embodiment of the present disclosure;

[0051] Figure 3 A flowchart of a method for optimizing dialogue information provided by another exemplary embodiment of the present disclosure;

[0052] Figure 4 A flowchart of a method for optimizing dialogue information provided by another exemplary embodiment of the present disclosure;

[0053] Figure 5 A flowchart of an intervention dialogue information flow provided for an exemplary embodiment of the present disclosure;

[0054] Figure 6 A structural diagram of a system for optimizing dialogue information provided by another exemplary embodiment of the present disclosure;

[0055] Figure 7 A schematic structural diagram of an electronic device provided by an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0056] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0057] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present disclosure are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0058] It should also be understood that in the embodiments of the present disclosure, “plurality” may refer to two or more than two, and “at least one” may refer to one, two, or more than two.

[0059] It should also be understood that any component, data or structure mentioned in the embodiments of the present disclosure can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0060] In addition, the term "and / or" in the present disclosure is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are in an "or" relationship.

[0061] It should also be understood that the description of the various embodiments in the present disclosure focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.

[0062] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0063] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0064] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0065] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0066] The disclosed embodiments can be applied to electronic devices such as terminal devices, computer systems, servers, etc., which can operate with many other general or special computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, servers, etc. include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems, and distributed cloud computing technology environments including any of the above systems, etc.

[0067] Electronic devices such as terminal devices, computer systems, servers, etc. can be described in the general context of computer system executable instructions (such as program modules) executed by computer systems. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0068] Prompt refers to text or instructions that provide input to the agent to guide the agent to generate specific outputs. It is used to describe the task goal to the agent so that the agent can output results that meet the task goal. For example, prompt can be a short question, a descriptive text, or a set of instruction information, which is used to provide guidance suggestions for the agent to process input information.

[0069] In the related art, in an LLM-based intelligent agent system, the intelligent agent iteratively processes tasks. However, during the iteration process, the task description information generated in multiple rounds may be completely consistent, and the output results of the intelligent agent cannot be optimized, and the intelligent agent falls into an invalid loop.

[0070] In the disclosed embodiment, by monitoring the historical dialogue information iteratively generated during the iterative task processing of the intelligent agent, intervening in the dialogue between the intelligent agents when similar dialogue information and / or invalid dialogue information exists in the historical dialogue information, and updating the task description information most recently generated by the intelligent agent, dynamic intervention in the iterative task processing process of the intelligent agent can be achieved, so that the dialogue between the intelligent agents develops in a convergent direction and avoids the intelligent agents from being in an invalid loop.

[0071] In a possible embodiment, the embodiment of the present disclosure can be applied to natural language processing technology, and specifically can be applied to intelligent question and answer scenarios, intelligent recommendation scenarios, automatic code generation scenarios, text translation scenarios, etc.

[0072] In one example, when applied to an intelligent question-answering scenario, the intelligent agent can receive input questions, iteratively process the input questions, and dynamically adjust the task description information that provides the intelligent agent with question processing suggestions during the iterative processing, so that the intelligent agent can output better answers to the questions.

[0073] In another example, when applied to a text translation scenario, the agent iteratively translates the input text to be translated, and dynamically adjusts the task description information that provides translation suggestions to the agent during the iteration process, so that the agent can output better translation results.

[0074] like Figure 1 As shown, a system for optimizing dialogue information provided by an exemplary embodiment of the present disclosure is shown. The system includes an adjustment module 11 and at least one intelligent agent 12.

[0075] At least one agent 12 processes tasks based on an iterative mechanism, processes user input information to obtain output information and outputs it to the user. Exemplarily, the input information may be at least one of input text information, input voice information, and input picture information.

[0076] When an intelligent agent is included, the intelligent agent can process tasks through a self-reflection mechanism. In the process of processing tasks based on the self-reflection mechanism, after receiving user input information, the first round of processing results are generated based on the initial prompt and the user input information. The initial prompt includes the role positioning of the intelligent agent, the description of its capabilities, the description of the user input information, etc. After the first round of processing results are generated, the intelligent agent proposes suggestion information on the first round of processing results based on the self-reflection mechanism, and inputs the suggestion information into the intelligent agent. The intelligent agent updates the initial prompt based on the suggestion information, and performs a second round of processing based on the updated prompt, and continuously iterates the task to obtain better task processing results.

[0077] When multiple agents are included, the multiple agents process tasks through collaborative iteration. Exemplarily, the multiple agents may include an execution agent and an evaluation agent, wherein the execution agent is an agent used to process tasks, and the evaluation agent is used to provide suggestion information for the task processing results output by the execution agent, so that the execution agent reprocesses the task according to the suggestion information. In an exemplary embodiment, after the execution agent receives input information, it generates a first round of processing results based on the input information and the initial prompt, and outputs it to the evaluation agent. The evaluation agent can generate suggestion information based on the first round of processing results, input the suggestion information to the execution agent, so that the execution agent updates the initial prompt based on the suggestion information, and enters the next round of iteration based on the updated prompt. That is, during the i-th round of iterations, the executing agent obtains the i-th round of processing results based on the i-th round of prompt processing tasks, inputs the i-th round of processing results to the evaluation agent, the evaluation agent proposes i-th suggestion information for the i-th round of processing results, inputs the i-th suggestion information to the executing agent, the executing agent updates the i-th round of prompt based on the i-th suggestion information to obtain the i+1-th round of prompt, and starts the i+1-th round of processing based on the i+1-th round of prompt.

[0078] The adjustment module 11 is used to dynamically intervene in the process of iteratively processing tasks by at least one agent 12. Exemplarily, the adjustment module 11 may be an agent different from the at least one agent 12, and may monitor the dialogue information during the iterative processing of tasks by the agent 12, and analyze the dialogue information. When the analysis shows that there is similar dialogue information and / or invalid dialogue information in the historical dialogue information, the last task description information input to the agent in the most recent round may be updated to adjust the task description information.

[0079] In a possible implementation, the adjustment module 11 is used to obtain historical dialogue information corresponding to the task in response to the end of a round of dialogue during the process of at least one agent processing the task, the historical dialogue information includes at least one round of dialogue information generated by at least one agent in the most recent round, and each round of dialogue information includes task description information input to at least one agent; perform content analysis on the historical dialogue information to obtain a content score of the historical dialogue information, and the content score is used to characterize at least one of the similarity and effectiveness of the historical dialogue information; in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, update the previous task description information input to at least one agent in the most recent round of dialogue, obtain updated task description information, and send the updated task description information to at least one agent. At least one agent 12 is used to receive the updated task description information and conduct the next round of dialogue based on the updated task description information.

[0080] In an illustrative example, when the adjustment module 11 dynamically intervenes in the iterative processing task of an agent, a round of dialogue information of an agent includes the agent outputting a round of processing results and suggestion information for the round of processing results.

[0081] In another illustrative example, when the adjustment module 11 dynamically intervenes in the collaborative iteration process of multiple agents, a round of dialogue information of the multiple agents includes a round of processing results output by the execution agent and suggestion information generated by the evaluation agent for the round of processing results.

[0082] In response to the end of a round of dialogue, the adjustment module 11 obtains at least one round of dialogue information generated in the process of the agent processing this task, that is, the historical dialogue information, and analyzes the historical dialogue information. If there is similar dialogue information and / or invalid dialogue information in the historical dialogue information, the task description information most recently input to the agent is updated to improve the quality of the content generated by the agent and make the agent dialogue develop in a convergent direction.

[0083] like Figure 2 As shown, it shows a flow chart of a method for optimizing dialogue information provided by an exemplary embodiment of the present disclosure. The method provided by the embodiment of the present disclosure can be used in the management module 11 described in the above embodiment, such as Figure 2 As shown, the following steps 201-203 are included:

[0084] Step 201, during the process of the intelligent agent processing this task, in response to the end of a round of dialogue, obtain the historical dialogue information corresponding to this task.

[0085] The historical dialogue information includes at least one round of dialogue information generated by the agent in at least one recent round, and each round of dialogue information includes task description information input to the agent. A round of dialogue information includes a round of processing results generated by the agent and task description information newly generated based on the round of processing results (i.e., the suggestion information described in the above embodiment).

[0086] In one possible implementation, after generating new task description information, it is determined that a round of dialogue ends, and after each round of dialogue ends, at least one round of dialogue information generated by the agent in at least one recent round can be obtained. Schematically, based on the first preset number of rounds, at least one round of dialogue information generated in at least one recent round can be obtained. For example, the first preset number of rounds can be 2, and two rounds of dialogue information generated in the most recent two rounds can be obtained. In another possible implementation, historical dialogue information can be periodically obtained, such as taking the second preset number of rounds as a period, and in response to the end of the last round of dialogue in a period, historical dialogue information including task description information generated in the period is obtained. Exemplarily, the second preset number of rounds can be 3.

[0087] Step 202: Perform content analysis on the historical conversation information to obtain content scores of the historical conversation information.

[0088] Content analysis is used to detect whether the task description information included in the dialogue information can effectively advance the progress of task processing. The content score obtained by content analysis can be used to characterize the degree to which the task description information included in the dialogue information advances the intelligent agent's task processing. Optionally, it can be used to characterize at least one of the similarity and effectiveness of historical dialogue information.

[0089] In a possible implementation, the content analysis includes detecting at least one of similarity between multiple rounds of dialogue information in the historical dialogue information and validity of the multiple rounds of dialogue information.

[0090] The similarity of multi-round dialogue information is negatively correlated with the content score. The higher the similarity, the higher the repetition of the agent's dialogue information and the lower the content score. The validity of the dialogue information is positively correlated with the content score. The higher the validity of the dialogue information, the higher the content score.

[0091] Step 203, in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, the previous task description information input to the agent in the most recent round of dialogue is updated to obtain updated task description information, so that the agent can conduct the next round of dialogue based on the updated task description information.

[0092] In a possible implementation, when the content score indicates that similar conversation information and / or invalid conversation information exists, the newly generated task description information in the most recent round of conversation may be updated to obtain updated task description information.

[0093] Optionally, the update content of the previous task description information may include at least one of guidance information for guiding the agent to modify the task description information, constraints on the previous task description information, and newly added task description information.

[0094] In an illustrative example, in response to the end of the i-th round of dialogue, historical dialogue information is obtained, content analysis is performed on the historical dialogue information to obtain a content score, and in response to the content score indicating that similar dialogue information and / or invalid dialogue information exists in the historical dialogue information, the newly generated task description information in the i-th round of dialogue is updated to obtain updated task description information, the updated task description information belongs to the i+1-th round of task description information, and the agent starts the next round of processing based on the i+1-th round of task description information to generate the next round of dialogue.

[0095] In a possible implementation, when the agent stops generating new task description information, the iterative processing process of the agent can be terminated, and the task processing result generated last time is output as the final result.

[0096] In the disclosed embodiment, during the process of the intelligent agent processing this task, the dialogue information iteratively generated by the intelligent agent is monitored. After a round of dialogue ends, at least one round of historical dialogue information generated by the intelligent agent in at least one recent round of processing this task is obtained. The historical dialogue information can be content detected to obtain at least one of the similarity and validity of the historical dialogue information. When similar dialogue information and / or invalid dialogue information are detected in the historical dialogue information, the previous task description information input to the intelligent agent in the most recent round of dialogue can be updated to provide different task description information for the intelligent agent to intervene in the iterative process of the intelligent agent, which helps to improve the quality of the content finally generated by the intelligent agent. Compared with the fixed round number limitation method in the related art, the disclosed embodiment can dynamically adjust the task description information input to the intelligent agent by monitoring and analyzing dialogue information of different rounds, which can guide the dialogue of the intelligent agent to develop in the direction of convergence, avoid the intelligent agent from falling into an invalid cycle, help to improve the convergence speed of the intelligent agent, improve the task processing efficiency, and reduce the waste of computing resources.

[0097] In addition, through the mechanism of dynamically monitoring dialogue information and adaptively adjusting task description information, the quality of the content generated by the intelligent agent and the dialogue convergence speed can be balanced. While ensuring the high quality of the generated content, the dialogue convergence speed of the intelligent agent can be improved, which helps to improve the intelligent agent's ability to handle complex tasks, achieve a balance between the quality of generated content and dialogue convergence, and help improve the stability of the system.

[0098] In a possible implementation, the quality score can be determined based on the similarity between multiple rounds of dialogue information, and the iterative processing of the agent can be dynamically intervened based on the quality score, such as Figure 3 As shown, the process includes steps 301-303:

[0099] Step 301, during the process of the intelligent agent processing this task, in response to the end of a round of dialogue, obtain the historical dialogue information corresponding to this task.

[0100] The implementation of step 301 may refer to the above step 201 and will not be described in detail in this embodiment.

[0101] Step 302: Perform similarity detection on the historical dialogue information to obtain a similarity score of the historical dialogue information. The similarity score is used to represent the similarity between dialogue information generated in different rounds.

[0102] In a possible implementation, similarity detection may be performed on at least two rounds of dialogue information to obtain a similarity score between the at least two rounds of dialogue information, wherein the at least two rounds of dialogue information include dialogue information generated in at least two recent rounds. Optionally, similarity detection may be performed based on a preset number of rounds, and after each preset number of rounds of dialogue information is obtained, similarity detection may be performed on the obtained dialogue information to obtain a similarity score. In an exemplary manner, the preset number of rounds may be 3, and after each three rounds of dialogue information generated in the iterative process are obtained, similarity detection may be performed on the dialogue information generated in the three iterative processes.

[0103] In a possible implementation, a task label included in each round of dialogue information in the historical dialogue information may be obtained, where the task label is used to indicate a processing task corresponding to the dialogue information; and a similarity score of the historical dialogue information may be determined based on the number of repetitions of the task label.

[0104] In each round of dialogue information generated by iteration, the task label of the task currently executed by the agent is carried, which is used to indicate the processing task targeted by the current dialogue information. Optionally, the task label can be the task label of the subtask that the agent is processing. When the agent receives input information, it will decompose the task into different subtasks (also called different steps) according to the task goal corresponding to the input information. Different subtasks correspond to different task labels. In the process of the agent executing each subtask, each dialogue information generated by iteration carries the task label of the subtask currently being processed.

[0105] The number of repetitions of a task label refers to the number of repetitions of the same task label. When the same task label is repeated multiple times, it indicates that the current agent has repeatedly performed the current subtask multiple times and may fall into a loop. The similarity score of the historical dialogue information can be determined based on the number of repetitions of the task label, wherein the number of repetitions of the task label is positively correlated with the similarity score. In an exemplary embodiment, a first correspondence between the number of repetitions and the similarity score can be preset, and the similarity score is determined based on the first correspondence and the number of repetitions. Exemplarily, if the repetition is less than 3 times, the similarity score can be determined to be 70; if the number of repetitions is greater than 3 times and less than 6 times, the similarity score can be determined to be 80; if the number of repetitions is not less than 6 times, the similarity score can be determined to be 90; if the number of repetitions is greater than or equal to 10 times, the similarity score can be determined to be 100.

[0106] In another possible implementation, the information content of different dialogue information in the historical dialogue information may be similarly detected to obtain a similarity score of the historical dialogue information. Optionally, the content similarity of different dialogue information in at least two rounds of dialogue information may be detected, and the similarity score of the historical dialogue information may be determined based on the similarity of the information content of different dialogue information. Exemplarily, the similarity between different task processing results included in at least two rounds of dialogue and the similarity between different task description information included in at least two rounds of dialogue may be detected respectively. Optionally, the similarity between the information contents may be determined based on at least one of the text similarity of the information content and the semantic similarity of the information content, wherein the text similarity and the semantic similarity are both positively correlated with the similarity. In an exemplary embodiment, the similarity ratio corresponding to the similarity degree corresponds to the similarity score. Exemplarily, the similarity ratio (1%-100%) corresponds to the similarity score (1-100). When the similarity degree is 70%, the similarity score is determined to be 70; when the similarity degree is 80%, the similarity score is determined to be 80.

[0107] In another possible implementation, a similarity score can be determined in combination with the number of repetitions of the task label and the similarity of the information content of the dialogue information to improve the accuracy of detecting the similarity of the task description information in the dialogue information. In response to the number of repetitions of the task label reaching a number threshold, a similarity detection is performed on the information content of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information. In this process, the number of repetitions of the task label can be detected first. When it is determined that the number of repetitions reaches a preset number threshold, it indicates that the agent may fall into multiple loops, and the similarity of the information content in different dialogue information is continuously detected. The similarity score is determined based on the similarity of the information content, so as to determine whether there is content increment (i.e., new task description information) to avoid invalid loops of the agent based on repeated task description information.

[0108] Illustratively, the number threshold may be 2. When two repetitions are detected, the similarity of different dialogue information in the historical dialogue information may continue to be detected.

[0109] Step 303, in response to the similarity score being higher than the similarity threshold, it is determined that similar dialogue information exists in the historical dialogue information, and the last task description information input to the agent in the most recent round of dialogue is updated.

[0110] When the similarity score is higher than the similarity threshold, it is determined that similar dialogue information exists in the historical dialogue information, and the previous task description information input to the agent in the most recent dialogue can be updated to obtain updated task description information. Among them, when the similarity score is higher than the similarity threshold, the first guidance information can be added to the previous task description information, and the first guidance information is used to indicate that the dialogue information of the agent is repeated, so that the agent that is iteratively processing the task modifies the task description information and develops in the direction of convergence.

[0111] Illustratively, the similarity threshold may be 70. When the similarity score is greater than 70, the previous task description information input to the agent in the most recent round of dialogue is updated.

[0112] In an illustrative example, when the similarity score is determined by combining the number of repetitions of the task label and the similarity of the information content, when the number of repetitions of the task label is greater than or equal to 2, the similarity between the contents of different dialogue information in the historical dialogue information can be detected. When the similarity is greater than 70%, it is determined that similar dialogue information exists in the historical dialogue information, and the previous task description information input to the agent in the most recent round of dialogue is updated.

[0113] In the disclosed embodiment, the similarity score of the historical dialogue information can be determined based on at least one of the number of repetitions of the historical dialogue information and the similarity of the information content. When the similarity score is higher than the similarity threshold, it is determined that there is similar, i.e., invalid repeated dialogue information, and the previous task description information can be updated to provide new and effective guidance information for the intelligent agent, thereby prompting the iterative dialogue between the intelligent agents to develop in a convergent direction.

[0114] In another possible implementation, the content score can be determined based on the validity of the information content contained in the historical dialogue information, and the iterative processing process of the agent can be dynamically intervened based on the content score, such as Figure 4 As shown, the process includes steps 401-403:

[0115] Step 401, during the process of the intelligent agent processing this task, in response to the end of a round of dialogue, obtain the historical dialogue information corresponding to this task.

[0116] The implementation of step 401 may refer to the above step 201 and will not be described in detail in this embodiment.

[0117] Step 402, perform validity check on the historical dialogue information to obtain a validity score of the historical dialogue information, where the validity score is used to characterize the degree of relevance between the dialogue information and the task goal to be processed by the agent.

[0118] In a possible implementation, the task description information included in the historical dialogue information can be tested for validity. The task description information may contain some information that is irrelevant to the task goal or has a low degree of relevance to the task goal. This information will affect the processing result of the intelligent agent. The validity score of the historical dialogue information can be obtained by testing the relevance of the task description information included in the historical dialogue information and the task goal to be processed by the intelligent agent. For the task description information included in a dialogue information, the different task descriptions contained in the task description information can be matched with the task goal, and the degree of relevance to the task goal is determined based on the degree of match with the task goal. Schematically, when the task goal is to generate a brief text describing the target scene, and the task description information contains multiple description suggestions that are irrelevant to the target scene or description suggestions for multiple details of the target scene, it can be determined that it does not match the task goal. In a possible implementation, the proportion of task descriptions that match the task goal included in the task description information can be determined as the degree of relevance between the task description information and the task goal.

[0119] In an exemplary embodiment, the correlation ratio corresponding to the degree of correlation between the task description information and the task objective corresponds to the effectiveness score. For example, the correlation ratio (1%-100%) corresponds to the effectiveness score (1-100). When the correlation degree is 80%, the effectiveness score is determined to be 80; when the correlation degree is 60%, the effectiveness score is determined to be 60.

[0120] Step 403, in response to the validity score being lower than the validity threshold, it is determined that invalid dialogue information exists in the historical dialogue information, and the last task description information input to the agent in the most recent round of dialogue is updated.

[0121] In one possible implementation, any dialogue information contained in the historical dialogue information may be checked for validity. When the validity score of any dialogue information is lower than a validity threshold, it is determined that invalid dialogue information exists in the historical dialogue information, and the previous task description information input to the agent in the most recent round of dialogue may be updated.

[0122] In another possible implementation, after performing validity detection on any dialogue information included in the historical dialogue information, the validity score of each dialogue information is obtained. When the proportion of dialogue information with validity scores lower than the validity value threshold to the total number of historical dialogue information is greater than a preset ratio threshold, it is determined that invalid dialogue information exists in the historical dialogue information, and the last task description information input to the intelligent agent in the most recent round of dialogue is updated. Indicatively, the preset ratio threshold may be 50%.

[0123] In response to the effectiveness score being lower than the effectiveness threshold, a second guidance information may be added to the previous task description information. The second guidance information is used to indicate that the agent task description information includes information irrelevant to the task objective, so that the agent can adjust the task description information according to the task objective, thereby improving the degree to which the task description information advances the processing task.

[0124] Exemplarily, the second guidance information may be determined based on the invalid information included in the task description information and the degree of modification of the iteratively generated result by the task description information. When the task description information includes multiple invalid information that are irrelevant to the task objective, the second guidance information may be determined as the task description information containing redundant information, and it is recommended to narrow the description scope and reduce redundant information. When the degree of modification of the iteratively generated result by the task description information is low, the second guidance information may include suggestions on the degree of modification, for example, the proposed task description information must modify the iterative result by 30%.

[0125] In the disclosed embodiments, the degree of correlation between the task description information included in the historical dialogue information and the task objective can be detected to obtain a validity score of the historical dialogue information. When the validity score indicates that the historical dialogue information includes task description information that is irrelevant to the task objective, it is determined that invalid dialogue information exists, and guidance information can be added to enable the intelligent agent to adjust the task description information, which helps to improve the degree of advancement of task processing and the quality of content generated by the intelligent agent, helps to balance the convergence speed of the intelligent agent and the quality of generated content, and improves the efficiency of task processing while improving the quality of generated content.

[0126] In one possible implementation, in response to a similarity score being higher than a similarity threshold, or in response to a validity score being lower than a validity threshold, the last task description information input to the agent in the most recent round of dialogue is updated. In another possible implementation, in response to a similarity score being higher than a similarity threshold and in response to a validity score being lower than a validity threshold, the last task description information input to the agent in the most recent round of dialogue is updated. Optionally, the first guidance information and the second guidance information may be added to the last task description information to obtain updated task description information, so that the agent develops in the direction of effective dialogue convergence.

[0127] In an exemplary embodiment, when updating the previous task description information input to the agent in the most recent round of dialogue, guidance information for guiding the agent to modify the task description information, constraints on the previous task description information, and at least one of newly added task description information can be added to the previous task description information.

[0128] Optionally, the guidance information includes the first guidance information and the second guidance information mentioned above, which can be determined based on the detection results of the similarity detection and the validity detection. On this basis, the constraints on the previous task description information can also be added, that is, the task description information generated by the current iteration round can be directly modified, and the constraints (such as the limiting conditions for different description information) can be added to avoid the invalid cycle of the agent, and the cycle can be performed again based on the modified task description information. In addition, the adjustment module is also an agent, and can also generate new task description information based on the current processing task. The newly added task description information is different from the task description information included in the historical dialogue information, and the task description information generated from different perspectives is provided to the agent to avoid the agent from falling into an invalid cycle.

[0129] In a possible implementation, the intervention intensity of the agent's iterative processing task can be determined based on the similarity level and / or invalidity level of the historical dialogue information. By dynamically adjusting the intervention intensity, the convergence speed of the agent can be further improved. Figure 5 As shown, in response to the content score indicating that the historical dialogue information has similar dialogue information and / or invalid dialogue information, updating the last task description information input to the agent in the most recent dialogue includes the following steps 501-503:

[0130] Step 501: in response to the content score indicating that the historical conversation information contains similar conversation information and / or invalid conversation information, determine the similarity level and / or invalidity level of the historical conversation information based on the content score.

[0131] If invalid conversation information is determined based on the validity score, the invalidity level of the historical conversation information is further determined based on the validity score. If similar conversation information is determined based on the similarity score, the similarity level of the historical conversation information is further determined based on the similarity score.

[0132] In a possible implementation, a first correspondence between the validity score and the invalidity level can be preset, and the invalidity level of the historical conversation information is determined according to the first correspondence and the validity score. The invalidity level is positively correlated with the indicated conversation invalidity level, and the higher the invalidity level, the higher the conversation invalidity level.

[0133] The validity score is negatively correlated with the invalidity level, that is, the lower the validity score, the higher the invalidity level. Indicatively, when the validity score is not greater than 70 and greater than 50, it is determined as the first invalidity level, when the validity score is not greater than 50 and greater than 20, it is determined as the second invalidity level, and when the similarity score is not greater than 20, it is determined as the third invalidity level.

[0134] In a possible implementation, a second correspondence between the similarity score and the similarity level may be preset, and the similarity level of the historical conversation information may be determined based on the second correspondence and the similarity score. The similarity level is positively correlated with the indicated conversation similarity. The higher the similarity level, the higher the conversation similarity.

[0135] The similarity score is positively correlated with the similarity level, that is, the higher the similarity score, the higher the similarity level. In schematic form, when the similarity score is not less than 70 and less than 80, it is determined as the first similarity level, when the similarity score is not less than 80 and less than 90, it is determined as the second similarity level, and when the similarity score is not less than 90, it is determined as the third similarity level.

[0136] Step 502: Determine the update content of the previous task description information based on the similarity level and / or the invalidity level.

[0137] The similarity level and / or invalidity level is positively correlated with at least one of the amount of information content contained in the update content and the intervention intensity of the update content on the agent's processing task.

[0138] When the similarity level and / or invalidity level is higher, the corresponding update content contains more information content, and the intervention intensity of the update content on the agent's processing task is higher.

[0139] In combination with the above examples, the first guidance information can be added at the first similarity level, and at the second similarity level, in addition to adding the first guidance information, at least one of the constraints on the previous task description information and the newly added task description information can be added. The second guidance information can be added at the first invalidity level, and at the second invalidity level, in addition to adding the second guidance information, at least one of the constraints on the previous task description information and the newly added task description information can be added.

[0140] When the update content is determined based on the similarity level and the invalidity level, the update content may be determined based on the highest level. Exemplarily, when the similarity level is the first and the invalidity level is the first, at least one of the first guide information and the second guide information may be added, and when the second similarity level and / or the second invalidity level exists, in addition to adding at least one of the first and the second guide information, at least one of the constraint conditions for the previous task description information and the newly added task description information may be added.

[0141] The intensity of intervention in the agent's task processing can be adjusted by adding the tone intensity corresponding to the updated content. The tone intensity refers to the intensity of the suggestions provided to the agent. Exemplarily, the tone intensity at the first similarity level and / or the first invalidity level can be a suggestion tone, suggesting that the agent modify the task description information based on the added updated content; when there is a second similarity level and / or a second invalidity level, the tone intensity is an execution tone, causing the agent to modify the task description information based on the updated content.

[0142] Step 503: Based on the updated content, the last task description information input to the agent in the most recent round of dialogue is updated.

[0143] After the update content is determined, the determined update content can be added to the previous task description information.

[0144] In one possible implementation, there may be a situation where the agent is already in an invalid loop. After determining the similarity level and / or invalidity level of the historical dialogue information based on the content score, in response to the similarity level belonging to a preset severe similarity level, a terminate dialogue instruction is sent to the agent to cause the agent to stop iterating; and / or, in response to the invalidity level belonging to a preset severe invalidity level, a terminate dialogue instruction is sent to the agent to cause the agent to stop iterating.

[0145] When it is determined that the similarity level and / or invalidity level belongs to the preset severity level, the iteration cycle of the agent can be stopped directly, and a termination dialogue instruction can be sent to the agent to save computing resources. In combination with the above example, the preset severity level can be the third level. When it is determined that the invalidity level belongs to the third invalidity level, a termination dialogue instruction is sent to the agent to stop the iteration of the agent, and the last generated processing result is output as the final target result. When it is determined that the similarity level belongs to the third similarity level, a termination dialogue instruction is sent to the agent to stop the iteration of the agent, and the last generated processing result is output as the final target result.

[0146] In the disclosed embodiment, the task description information is adjusted in real time in combination with at least one of the similarity and effectiveness of the historical dialogue information, and additional information or suggestions are added to ensure that each round of dialogue can provide a new perspective for the final solution. In addition, the intervention intensity of the task description information can be dynamically changed according to the similarity level and / or invalidity level. As the similarity level and / or invalidity level increases, the intervention intensity can be gradually increased to ensure that the dialogue of the intelligent agent converges in a shorter time. The adaptive adjustment mechanism can effectively improve the processing capacity of complex tasks and improve the efficiency of the intelligent agent in iteratively processing tasks while ensuring the quality of the content.

[0147] The disclosed embodiment, by introducing an intelligent dynamic monitoring and intervention mechanism, can effectively solve the task with a relatively small number of interactions, avoid unnecessary repeated iterations of the agent, and reduce the waste of computing resources and operating costs. Compared with the round limit method in the related art, the disclosed embodiment can significantly reduce the number of interactions of the agent by adjusting the dialogue information at the right time to guide the dialogue convergence, saving resources while ensuring the quality and diversity of the generated results.

[0148] In a possible implementation, Figure 6 As shown, the dialogue information optimization system includes an adjustment module 11 including a dialogue monitoring unit 601, a content analysis unit 602 and a dialogue adjustment unit 603, wherein the dialogue monitoring unit 601 can be used to obtain historical dialogue information generated by the intelligent agent in the process of processing this task, the content analysis unit 602 performs content analysis on the obtained historical dialogue information to obtain a content score, and the dialogue adjustment unit 603 dynamically adjusts the dialogue of the intelligent agent according to the content score.

[0149] In an exemplary embodiment, a dialogue monitoring unit 601 is used to obtain historical dialogue information corresponding to the task in response to the end of a round of dialogue during the process of the intelligent agent processing the task, wherein the historical dialogue information includes at least one round of dialogue information generated by the intelligent agent in at least one recent round, and each round of dialogue information includes task description information input to the intelligent agent; a content analysis unit 602 is used to perform content analysis on the historical dialogue information to obtain a content score of the historical dialogue information, and the content score is used to characterize at least one of the similarity and validity of the historical dialogue information; a dialogue adjustment unit 603 is used to update the previous task description information input to the intelligent agent in the most recent round of dialogue in response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, to obtain updated task description information, so that the intelligent agent can conduct the next round of dialogue based on the updated task description information.

[0150] In an exemplary embodiment, the content analysis unit 602 is further configured to:

[0151] Performing a similarity test on the historical dialogue information to obtain a similarity score for the historical dialogue information, wherein the similarity score is used to represent the degree of similarity between dialogue information generated in different rounds;

[0152] and / or,

[0153] The historical dialogue information is tested for validity to obtain a validity score for the historical dialogue information, where the validity score is used to characterize the degree of relevance between the dialogue information and the task objective to be processed by the agent.

[0154] In an exemplary embodiment, the content analysis unit 602 is further configured to:

[0155] Obtaining a task tag included in each round of dialogue information in the historical dialogue information, wherein the task tag is used to indicate a processing task corresponding to the dialogue information; and determining a similarity score of the historical dialogue information based on the number of repetitions of the task tag;

[0156] and / or,

[0157] A similarity test is performed on the information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

[0158] In an exemplary embodiment, the content analysis unit 602 is further configured to:

[0159] In response to the repetition number of the task tag reaching a number threshold, similarity detection is performed on information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

[0160] In an exemplary embodiment, the dialogue adjustment unit 603 is further configured to:

[0161] In response to the similarity score being higher than a similarity threshold, determining that the historical dialogue information contains the similar dialogue information, and updating the last task description information input to the agent in the most recent round of dialogue;

[0162] and / or,

[0163] In response to the validity score being lower than a validity threshold, it is determined that the historical dialogue information contains the invalid dialogue information, and the last task description information input to the agent in the most recent round of dialogue is updated.

[0164] In an exemplary embodiment, the dialogue adjustment unit 603 is further configured to:

[0165] In response to the similarity score being higher than a similarity threshold, adding first guidance information to the previous task description information, wherein the first guidance information is used to indicate that the agent dialogue information is repeated;

[0166] In response to the effectiveness score being lower than the effectiveness threshold, second guidance information is added to the previous task description information, wherein the second guidance information is used to indicate that the agent task description information includes information that is irrelevant to the task goal.

[0167] In an exemplary embodiment, the dialogue adjustment unit 603 is further configured to:

[0168] In response to the content score indicating that similar conversation information and / or invalid conversation information exists in the historical conversation information, determining a similarity level and / or an invalidity level of the historical conversation information based on the content score;

[0169] Determining update content for updating the previous task description information based on the similarity level and / or the invalidity level, wherein the similarity level and / or the invalidity level is positively correlated with at least one of the amount of information content contained in the update content and the intensity of intervention of the update content on the task processed by the agent;

[0170] Based on the updated content, the last task description information input to the agent in the most recent round of dialogue is updated.

[0171] In an exemplary embodiment, the dialogue adjustment unit 603 is further configured to:

[0172] In response to the similarity level belonging to a preset severe similarity level, sending a dialogue termination instruction to the agent so that the agent stops iterating;

[0173] and / or,

[0174] In response to the invalidity level belonging to the preset serious invalidity level, a dialogue termination instruction is sent to the agent to stop the iteration of the agent.

[0175] In an exemplary embodiment, the dialogue adjustment unit 603 is further configured to:

[0176] In the previous task description information, at least one of guidance information for guiding the intelligent agent to modify the task description information, constraint conditions on the previous task description information, and newly added task description information is added.

[0177] The dialogue information optimization system in the embodiment of the present disclosure corresponds to the embodiment of the dialogue information optimization method in the present disclosure, and the relevant contents can be referenced to each other, which will not be repeated here. The beneficial technical effects corresponding to the dialogue information optimization device in the embodiment of the present disclosure can refer to the corresponding beneficial technical effects of the corresponding exemplary method part above, which will not be repeated here.

[0178] In addition, an embodiment of the present disclosure further provides an electronic device, including:

[0179] Memory for storing computer programs;

[0180] The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the method for optimizing the dialogue information described in any of the above embodiments of the present disclosure is implemented.

[0181] Figure 7This is a schematic diagram of the structure of an application embodiment of the electronic device disclosed in the present invention. Figure 7 To describe the electronic device according to the embodiment of the present disclosure. Figure 7 As shown, the electronic device includes one or more processors and memory.

[0182] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0183] The memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement the above-described method for optimizing the dialog information of each embodiment of the present disclosure and / or other desired functions.

[0184] In one example, the electronic device may further include: an input device and an output device, and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0185] In addition, the input device may also include, for example, a keyboard, a mouse, and the like.

[0186] The output device can output various information to the outside, including the determined distance information, direction information, etc. The output device can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0187] Of course, to simplify, Figure 7 Only some of the components related to the present disclosure in the electronic device are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device may further include any other appropriate components.

[0188] In addition to the above-mentioned methods and devices, an embodiment of the present disclosure may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method for optimizing dialogue information according to various embodiments of the present disclosure described in the above part of this specification.

[0189] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the disclosed embodiments, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0190] In addition, an embodiment of the present disclosure may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, enables the processor to execute the steps of the method for optimizing conversation information according to various embodiments of the present disclosure described in the above part of this specification.

[0191] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0192] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes.

[0193] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limiting. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0194] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0195] The block diagrams of the devices, apparatuses, equipment, and systems involved in this disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open words, referring to "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0196] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0197] It should also be noted that in the apparatus, device and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0198] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0199] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for optimizing dialogue information, characterized in that: The method comprises: In the process of the agent processing the task, in response to the end of a round of dialogue, historical dialogue information corresponding to the task is obtained, wherein the historical dialogue information includes at least one round of dialogue information generated by the agent in at least one recent round, and each round of dialogue information includes task description information input to the agent; Performing content analysis on the historical conversation information to obtain a content score of the historical conversation information, wherein the content score is used to characterize at least one of similarity and effectiveness of the historical conversation information; In response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, the previous task description information input to the agent in the most recent round of dialogue is updated to obtain updated task description information, so that the agent conducts the next round of dialogue based on the updated task description information.

2. The method according to claim 1, characterized in that The performing content analysis on the historical conversation information to obtain a content score of the historical conversation information includes: Performing a similarity test on the historical dialogue information to obtain a similarity score for the historical dialogue information, wherein the similarity score is used to represent the degree of similarity between dialogue information generated in different rounds; and / or, The historical dialogue information is tested for validity to obtain a validity score for the historical dialogue information, where the validity score is used to characterize the degree of relevance between the dialogue information and the task objective to be processed by the agent.

3. The method according to claim 2, characterized in that The performing similarity detection on the historical conversation information to obtain a similarity score of the historical conversation information includes: Obtaining a task tag included in each round of dialogue information in the historical dialogue information, wherein the task tag is used to indicate a processing task corresponding to the dialogue information; and determining a similarity score of the historical dialogue information based on the number of repetitions of the task tag; and / or, A similarity test is performed on the information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

4. The method according to claim 3, characterized in that After obtaining the task label included in each round of dialogue information in the historical dialogue information, the method further includes: In response to the repetition number of the task tag reaching a number threshold, similarity detection is performed on information contents of different dialogue information in the historical dialogue information to obtain a similarity score of the historical dialogue information.

5. The method according to claim 2, characterized in that: In response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, updating the last task description information input to the agent in the most recent round of dialogue, including: In response to the similarity score being higher than a similarity threshold, determining that the historical dialogue information contains the similar dialogue information, and updating the last task description information input to the agent in the most recent round of dialogue; and / or, In response to the validity score being lower than a validity threshold, it is determined that the historical dialogue information contains the invalid dialogue information, and the last task description information input to the agent in the most recent round of dialogue is updated.

6. The method according to claim 5, characterized in that The updating of the last task description information input to the agent in the most recent round of dialogue includes: In response to the similarity score being higher than a similarity threshold, adding first guidance information to the previous task description information, wherein the first guidance information is used to indicate that the agent dialogue information is repeated; In response to the effectiveness score being lower than the effectiveness threshold, second guidance information is added to the previous task description information, wherein the second guidance information is used to indicate that the agent task description information includes information that is irrelevant to the task goal.

7. The method according to any one of claims 1 to 6, characterized in that: In response to the content score indicating that the historical dialogue information contains similar dialogue information and / or invalid dialogue information, updating the last task description information input to the agent in the most recent round of dialogue, including: In response to the content score indicating that similar conversation information and / or invalid conversation information exists in the historical conversation information, determining a similarity level and / or invalidity level of the historical conversation information based on the content score; Determining update content for updating the previous task description information based on the similarity level and / or the invalidity level, wherein the similarity level and / or the invalidity level is positively correlated with at least one of the amount of information content contained in the update content and the intensity of intervention of the update content on the task processed by the agent; Based on the updated content, the last task description information input to the agent in the most recent round of dialogue is updated.

8. The method according to claim 7, characterized in that After determining the similarity level and / or invalidity level of the historical conversation information based on the content score, the method further includes: In response to the similarity level belonging to a preset severe similarity level, sending a dialogue termination instruction to the agent so that the agent stops iterating; and / or, In response to the dialogue invalidity level belonging to a preset serious invalidity level, a dialogue termination instruction is sent to the agent to stop the iteration of the agent.

9. An electronic device, characterized in that: include: Memory for storing computer programs; The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the method for optimizing the dialogue information described in any one of claims 1 to 8 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for optimizing the dialogue information described in any one of claims 1 to 8 is implemented.

Images