Dialogue processing method and apparatus, and electronic device and storage medium

Abstract

Provided are a dialogue processing method and apparatus, and an electronic device, a storage medium, a computer program product and a computer program, which relate to the technical field of computers and specifically relate to the technical fields of artificial intelligence such as large language models, natural language processing, knowledge graphs, and deep learning. The specific implementation scheme comprises: acquiring a topic of the current dialogue of a first user and a first historical dialogue library associated with the first user; determining the similarity between the topic of the current dialogue and each historical dialogue in the first historical dialogue library; on the basis of the similarities, acquiring a plurality of candidate historical dialogues from the first historical dialogue library; on the basis of the forgetting coefficient and weight of each candidate historical dialogue, acquiring a reference dialogue from the plurality of candidate historical dialogues; and generating a reply statement on the basis of the reference dialogue and the topic of the current dialogue.

Description

Dialogue processing method, device, electronic device and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Chinese Patent Application No. 2024101715645 filed in China on February 6, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The embodiments of the present disclosure relate to the field of computer technology, and in particular to the field of artificial intelligence technologies such as large language models, natural language processing, knowledge graphs, and deep learning, and specifically to a dialogue processing method and apparatus, electronic equipment, storage medium, computer program product, and computer program. Background Art

[0004] With the development of artificial intelligence (AI) technology, large language models and their applications have garnered widespread attention. Therefore, improving the personalization and accuracy of responses generated by large language models during conversations has become a pressing issue, both for enhancing user experience and increasing application competitiveness.

[0005] Summary of the Invention

[0006] The embodiments of the present disclosure aim to solve one of the technical problems in the related art at least to a certain extent.

[0007] To this end, the purpose of the embodiments of the present disclosure is to propose a conversation processing method and device, electronic device, storage medium, computer program product and computer program, which can improve the reliability and personalization of the generated reply statements, improve the quality of conversation, and increase user chat satisfaction.

[0008] According to an embodiment of the first aspect of the present disclosure, a method for processing a conversation is provided, including:

[0009] Obtaining a topic of a current conversation with a first user and a first historical conversation library associated with the first user;

[0010] Determining the similarity between the topic of the current conversation and each historical conversation in the first historical conversation library;

[0011] According to the similarity, a plurality of candidate historical conversations are obtained from the first historical conversation library;

[0012] Obtaining a reference dialogue from the plurality of candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue;

[0013] A reply statement is generated based on the reference conversation and the subject of the current conversation.

[0014] According to a second embodiment of the present disclosure, there is provided a dialog processing device, comprising:

[0015] A first acquisition module is configured to acquire a topic of a current conversation with a first user and a first historical conversation library associated with the first user;

[0016] a first determining module, configured to determine a similarity between the topic of the current conversation and each historical conversation in the first historical conversation library;

[0017] A second acquisition module is configured to acquire a plurality of candidate historical conversations from the first historical conversation library based on the similarity;

[0018] A third acquisition module is configured to acquire a reference dialogue from the plurality of candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue;

[0019] A generation module is used to generate a reply statement based on the reference dialogue and the subject of the current dialogue.

[0020] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0021] at least one processor; and

[0022] a memory communicatively connected to the at least one processor; wherein,

[0023] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the dialogue processing method as described in the first aspect.

[0024] According to an embodiment of the fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the dialogue processing method as described in the first aspect.

[0025] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising computer instructions, which, when executed by a processor, implement the steps of the dialog processing method as described in the first aspect.

[0026] According to a sixth aspect of the present disclosure, a computer program is provided, comprising computer program code. When the computer program code is run on a computer, the computer executes the dialog processing method as described in the first aspect.

[0027] The conversation processing method and apparatus, electronic device, storage medium, computer program product, and computer program provided by the embodiments of the present disclosure have the following beneficial effects:

[0028] In the disclosed embodiment, the topic of a first user's current conversation and the first user's associated historical conversation database are first obtained. The similarity between the topic of the current conversation and each historical conversation in the first historical conversation database is then determined. Based on the similarity, multiple candidate historical conversations are retrieved from the first historical conversation database. Reference conversations are then retrieved from the multiple candidate historical conversations based on the forgetting coefficient and weight of each candidate historical conversation. Finally, a reply statement is generated based on the reference conversations and the topic of the current conversation. Thus, historical statements that are similar to the topic of the current conversation and highly timely are screened from the user's associated historical conversation database as reference information to assist in generating a reply statement. This improves the reliability and personalization of the reply statement, enhances the quality of the conversation, and increases user chat satisfaction.

[0029] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and / or additional aspects and advantages of the present disclosure will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, which are provided for a better understanding of the present solution and do not constitute a limitation of the present disclosure.

[0031] FIG1 is a flow chart of a method for processing a conversation according to an embodiment of the present disclosure;

[0032] FIG2 is a schematic diagram of the structure of a user portrait map provided by an embodiment of the present disclosure;

[0033] FIG3 is a flow chart of a conversation processing method according to another embodiment of the present disclosure;

[0034] FIG4 is a flow chart of a method for processing a conversation according to another embodiment of the present disclosure;

[0035] FIG5 is a schematic diagram of the structure of an updated user portrait map provided by an embodiment of the present disclosure

[0036] FIG6 is a flow chart of a conversation processing method according to another embodiment of the present disclosure;

[0037] FIG7 is a structural diagram of a conversation processing device according to an embodiment of the present disclosure;

[0038] FIG8 illustrates a block diagram of an exemplary electronic device suitable for implementing embodiments of the present disclosure. DETAILED DESCRIPTION

[0039] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0040] The embodiments of the present disclosure relate to artificial intelligence technology fields such as large language models, natural language understanding, knowledge graphs, and deep learning.

[0041] Artificial Intelligence (AI) is a new technical science that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence.

[0042] Large Language Models (LLMs) are deep learning models trained on massive amounts of text data. They can not only generate natural language text but also deeply understand the meaning of text and handle various natural language tasks such as text summarization, question answering, and translation.

[0043] Natural language understanding (NLU), commonly known as human-computer dialogue, is a sub-discipline of artificial intelligence. It studies the use of computers to simulate human language communication processes, enabling computers to understand and use natural languages ​​such as Chinese and English, enabling natural language communication between humans and machines, thereby replacing some of the human mental work involved in searching for information, answering questions, extracting literature, compiling materials, and processing all natural language information.

[0044] Knowledge Graph, known as knowledge domain visualization or knowledge domain mapping map in the library and information industry, is a series of various graphs that show the development process and structural relationship of knowledge. It uses visualization technology to describe knowledge resources and their carriers, and to mine, analyze, construct, draw and display knowledge and their mutual connections.

[0045] Deep learning involves learning the inherent patterns and representational hierarchies of sample data. The information gained from this learning process is highly helpful in interpreting data such as text, images, and sounds. The ultimate goal of deep learning is to enable machines to have the same analytical and learning capabilities as humans, enabling them to recognize data such as text, images, and sounds.

[0046] In the technical solutions of the embodiments of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0047] The following describes the dialog processing method and apparatus, electronic device, storage medium, computer program product, and computer program according to embodiments of the present disclosure with reference to the accompanying drawings.

[0048] It should be noted that the execution subject of the dialogue processing method of this embodiment is a dialogue processing device, which can be implemented by software and / or hardware. The device can be configured in an electronic device, and the electronic device may include but is not limited to a terminal, a server, etc.

[0049] In the embodiment of the present disclosure, the dialogue processing device can be configured in a dialogue system of any application or website, etc. The dialogue system can generate reply content and conduct dialogue with the user through the dialogue processing method provided in the embodiment of the present disclosure.

[0050] FIG1 is a flow chart of a conversation processing method according to an embodiment of the present disclosure.

[0051] As shown in FIG1 , the dialogue processing method includes steps S101 to S105 .

[0052] S101: Obtain a topic of a current conversation of a first user and a first historical conversation library associated with the first user.

[0053] The first user may be any user who triggers the dialogue system. Triggering the dialogue system may include clicking to enter a chat interface of the dialogue system, or clicking a specific control in a program or website with a dialogue function, etc. This may be pre-set based on actual application needs and is not limited in the present embodiment.

[0054] In an embodiment of the present disclosure, when the current conversation interface contains the conversation content between the first user and the system, the conversation content in the interface can be analyzed and summarized to obtain the topic of the current conversation.

[0055] It should be noted that in the disclosed embodiments, the current conversation can be initiated not only by the first user but also by the conversation system. That is, if the first user triggers the conversation system but there is no conversation content in the current conversation interface, the conversation system can analyze the first user's behavior habits based on a pre-built user profile graph, determine the topic that the first user is most likely to be interested in, and use it as the topic of the current conversation, and then initiate a conversation based on this topic.

[0056] The following describes how the user portrait map is constructed with reference to FIG2 , which is a schematic diagram of the structure of the user portrait map.

[0057] In the disclosed embodiments, the behavioral data of each user of the program or website used by the dialogue system can be collected offline or in other ways, and behavioral tags associated with each user can be extracted from the behavioral data, such as the forum name, points of interest, consumption category, post tags, etc. For example, after collecting the behavioral habits of user a and user b and extracting behavioral tags for the behavioral habits of user a and user b respectively, the behavioral tags associated with user a can be obtained, including point of interest a, point of interest b, consumption category b, and post content corresponding to tag1 and tag2. The behavioral tags associated with user b include hot posts a and hot topics b in forum a, point of interest b, consumption category b, and post content corresponding to tags3, tag4, and tag5.

[0058] Next, user IDs a and b and their associated behavior tags are used as nodes in the user portrait graph. Based on the user's click, consumption, and other behaviors, the connection between the user ID nodes and the behavior tag nodes is determined, thereby constructing the user portrait graph, as shown in Figure 2. As shown in Figure 2, the vector representation of the user ID node for each behavior tag node can also be obtained based on the weights determined by the number of clicks on points of interest, the number of consumption in consumption categories, the length of time users browse post content or forums, and behaviors such as likes.

[0059] In the disclosed embodiment, the similarity between the vectors of the first user and the behavior tag can be calculated based on the vector representations of each node in the user portrait graph. Multiple candidate topics are then determined based on the order of similarity from largest to smallest. The multiple candidate topics are then sorted and screened based on the latest hot topics and the user's latest behavioral habits obtained through channels such as forums. The candidate topic with the highest similarity is determined as the topic that the first user is most likely to be interested in.

[0060] In the disclosed embodiment, the historical conversation library associated with each user may be pre-built based on the historical conversation content associated with the user. During a conversation, the first historical conversation library associated with the first user may be found based on the identifier of the first user (such as user name, account number, etc.).

[0061] S102: Determine the similarity between the topic of the current conversation and each historical conversation in the first historical conversation database.

[0062] It should be noted that the first historical conversation library may include summaries, keywords, and the like corresponding to each historical conversation. Therefore, when calculating the similarity between the topic of the current conversation and each historical conversation in the first historical conversation library, the similarity between the topic of the current conversation and the keywords of each historical conversation, or the similarity between the topic of the current conversation and the summaries of each historical conversation, etc., may be calculated, and this is not limited in the present embodiment.

[0063] S103: Acquire multiple candidate historical conversations from the first historical conversation database based on the similarity.

[0064] In the embodiment of the present disclosure, the first N corresponding historical conversations can be determined as candidate historical conversations in descending order of similarity, where N can be any positive integer, such as 5, 10, etc., and is not limited in the embodiment of the present disclosure.

[0065] S104: Based on the forgetting coefficient and weight of each candidate historical dialogue, a reference dialogue is obtained from multiple candidate historical dialogues.

[0066] The forgetting coefficient indicates the degree to which each historical conversation influences the current response. This coefficient is determined based on the user's emotional fluctuations in the historical conversations and the time interval between the corresponding conversations and the current time. The greater the user's emotional fluctuations and the smaller the time interval, the greater the forgetting coefficient, and the greater the impact on the current response.

[0067] It should be noted that the forgetting coefficient is updated over time. The longer the interval between the time when the historical conversation was generated and the current time, the smaller the forgetting coefficient. When the forgetting coefficient of a historical conversation is less than a certain value, the historical conversation can be deleted from the first historical conversation library.

[0068] Among them, the weight of the candidate historical conversation can be a value determined based on the frequency of occurrence of keywords that express the intention of the conversation in the historical conversation, the degree of emotional fluctuation of the user during the conversation, etc., which is used to describe the importance of each historical conversation in the historical conversation library.

[0069] It should be noted that when historical conversations are stored in the first historical database, their weights and forgetting rates are determined and stored in association with the historical conversations. Therefore, when selecting a reference conversation, the forgetting coefficient and weight of each candidate historical conversation can be directly obtained from the first historical database.

[0070] In some embodiments, at least one historical conversation with the largest forgetting coefficient and / or the highest weight among multiple candidate historical conversations can be identified. Then, the historical conversation with the smallest time interval between the generation time and the current time, which is less than a time threshold, can be determined as the reference conversation.

[0071] It should be noted that when the time interval between the generation time of the historical conversation and the current time is less than the time threshold, the historical conversation is highly timely and may have good reference value for generating a reply statement. The time threshold can be determined based on business needs, for example, 10 days, 30 days, etc., and is not limited in this embodiment of the present disclosure.

[0072] In the disclosed embodiment, by determining the reference dialogues to assist in generating the reply statements based on the forgetting coefficient, weight and generation time of the historical dialogues, it is possible to ensure that the screened reference dialogues are more reliable and have more reference value, thereby promoting the improvement of the quality and personalization of the reply statements.

[0073] S105, generating a reply statement based on the reference conversation and the subject of the current conversation.

[0074] In the disclosed embodiment, the reference conversation and the subject of the current conversation may be input into a language model of the conversation system, and a reply statement may be generated by the language model.

[0075] It should be noted that the language model can be a large language model (LLM), such as GPT, Wenxinyiyan, etc., or it can be other language models that can generate reply sentences. The embodiments of the present disclosure are not limited to this.

[0076] In some embodiments, when the time interval between the generation time of each historical conversation in at least one historical conversation and the current time is greater than a time threshold, a reply statement can be generated based on the topic of the current conversation.

[0077] It's understandable that when the time interval between a historical conversation and the current time exceeds a threshold, the historical conversation may not accurately reflect the user's current conversation preferences, potentially leading to deviations and semantic errors in the reply. Therefore, when generating a reply, it's possible to bypass historical conversation information and directly generate a reply relevant to the current conversation topic. This avoids using historical conversations from long ago as reference information for reply generation, further ensuring the accuracy and timeliness of the reply.

[0078] In this embodiment, the topic of the first user's current conversation and the first user's associated historical conversation database are first obtained. The similarity between the topic of the current conversation and each historical conversation in the first historical conversation database is determined. Based on the similarity, multiple candidate historical conversations are retrieved from the first historical conversation database. Reference conversations are then obtained from the candidate historical conversations based on the forgetting coefficient and weight of each candidate historical conversation. Finally, a reply is generated based on the reference conversations and the topic of the current conversation. Thus, historical conversations that are similar to the topic of the current conversation and highly timely are screened from the user's associated historical conversation database as reference information to assist in the generation of the reply. This improves the reliability and personalization of the reply, enhances the quality of the conversation, and increases user chat satisfaction.

[0079] FIG3 is a flow chart of a conversation processing method according to another embodiment of the present disclosure.

[0080] As shown in FIG3 , the dialogue processing method includes steps S301 to S306 .

[0081] S301: Obtain a topic of a current conversation of a first user and a first historical conversation library associated with the first user.

[0082] S302: Determine the similarity between the topic of the current conversation and each historical conversation in the first historical conversation database.

[0083] S303: Acquire multiple candidate historical conversations from the first historical conversation database based on the similarity.

[0084] S304: Obtain a reference dialogue from multiple candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue.

[0085] The description of the above S301 to S304 can be found in the above embodiments and will not be repeated here.

[0086] It's important to note that in some application scenarios, such as gaming and cultural creation, the dialogue system may need to simulate the speech characteristics of a specific character to engage in conversation with the user. Therefore, before generating a response, it's necessary to determine the target character to simulate.

[0087] S305: Determine the role type of the dialogue system in the current dialogue.

[0088] The role type may be classified based on at least one of gender, age range, occupation, speaking tone or habits, character relationships, etc. Alternatively, the roles in the dialogue system may also include certain specific roles set by movies, games, etc., which is not limited in the present embodiment.

[0089] It's important to note that the role type the dialogue system plays during a conversation is fixed and determined before each conversation begins. The role type the dialogue system plays during a conversation can be selected by the first user based on their needs, or based on the first user's historical preferences. Once the selected role type is stored in the system, it will be used each time a response is generated.

[0090] In some embodiments, when the first user has not selected a role type, the dialogue system can determine the role type with the highest chat frequency as the role type in the current dialogue based on the role types used historically by the first user and the chat frequency with each role type.

[0091] In some embodiments, when the first user has no history of using a character type, the character type in the current conversation may also be determined based on the character types used historically by a second user similar to the first user and the chat frequency with each character type.

[0092] In the embodiment of the present disclosure, by determining the character type with the highest historical usage frequency as the target character type for the current conversation, the user's preferences and needs can be better met, the time cost of character selection can be reduced, and user satisfaction can be improved.

[0093] In some embodiments, a second user similar to the first user can be determined based on the behavioral tags associated with each user in the user portrait graph.

[0094] It's understandable that if different users have similar behavioral habits, they're likely to like the same type of characters. Therefore, we can first identify candidate topics corresponding to the first user based on the behavioral tags associated with them in the user portrait graph. Then, we can look at the behavioral tags of other users in the graph, match those tags with the candidate topics, and identify the user with the highest similarity as the second user similar to the first user.

[0095] Therefore, by determining the second user who is similar to the first user through the behavioral tags associated with each user in the user portrait graph, the accuracy and reliability of the second user determination can be improved, providing conditions for improving the accuracy of determining the role type preferred by the first user.

[0096] It should be noted that the role type most recently used by the first user or the second user may also be determined as the role type of the dialogue system in the current dialogue, which is not limited in this embodiment of the present disclosure.

[0097] S306, generating a reply statement based on the reference dialogue, the subject of the current dialogue and the prompt information associated with the role type.

[0098] The role-type-related prompts may include descriptions of the role type's gender, age range, occupation, voice tone or habits, and relationships with the characters. For example, the prompts could be "middle-aged female, teacher, strict," or "male, father, kind."

[0099] In this embodiment, after determining the reference dialogue for sentence generation, the dialogue system first determines the role type in the current dialogue. Then, based on the reference dialogue, the current dialogue topic, and the prompt information associated with the role type, a response sentence is generated. By generating a personalized response sentence based on the prompt sentence corresponding to the target role type, the diversity and quality of the response sentences can be further improved, enhancing the user's sense of interaction.

[0100] FIG4 is a flow chart of a conversation processing method according to another embodiment of the present disclosure.

[0101] As shown in FIG4 , the dialogue processing method includes steps S401 to S403 .

[0102] S401, obtaining incremental user data in the current cycle.

[0103] The incremental user data includes user IDs and associated behavior tags.

[0104] Understandably, user profiles are constructed based on historical user behavior data. However, new user behavior data is constantly generated, and the influence of historical data on analyzing user preferences decreases over time. Therefore, the effectiveness of previously constructed user profiles in determining target users' conversation topic preferences may be reduced. Therefore, user profiles need to be regularly updated to ensure that they are current and complete, thereby ensuring the accuracy of conversation topics determined by the dialogue system.

[0105] In the embodiment of the present disclosure, an update cycle can be preset, and behavior tags can be extracted from the newly generated user behavior data in the current cycle, associated with the user identifier, and used as user incremental data to update the user portrait map.

[0106] It should be noted that the update cycle can be determined according to actual conditions. For example, if the amount of newly added user behavior data is large, the cycle can be shortened, which is not limited in the present embodiment.

[0107] S402: Based on the user identification and behavior tags, the current user portrait graph is traversed.

[0108] In the disclosed embodiment, the current user profile graph can first be searched for each user ID in the incremental user data to determine whether the user ID is included. Then, based on the behavior tag associated with the user ID in the incremental user data, it is determined whether the behavior tag associated with the user ID in the current user profile graph is the same as that associated with the user ID.

[0109] S403: When the current user portrait graph contains the user identifier and does not contain at least one behavior tag in the user incremental data, the behavior tags in the current user portrait graph are updated based on the at least one behavior tag to obtain an updated user portrait graph.

[0110] In an embodiment of the present disclosure, when the user portrait graph contains any user identifier in the user incremental data, the behavior tag associated with the user identifier can be searched in the current user portrait graph to see whether it corresponds to the behavior tag associated with the user identifier in the user incremental data. If the current user portrait graph does not contain at least one behavior tag in the user incremental data, the non-contained behavior tag can be added to the current user portrait graph as a neighbor node of the user identifier, and a connection relationship can be established with the user identifier. Then, an aggregation function is used to train the newly added nodes and edges in the user portrait graph to complete the update of the user portrait graph.

[0111] In some embodiments, when the current user portrait map does not contain any user identifier in the user incremental data, the current user portrait map is updated based on any user identifier and the behavior tag associated with any user identifier to obtain an updated user portrait map.

[0112] In the disclosed embodiment, when the user portrait graph does not contain any user identifier in the user incremental data, a local graph can be constructed based on any user identifier and the behavioral label associated with any user identifier. Then, an aggregation function is used to merge the local graph with the current user portrait graph to complete the update of the user portrait graph. Thus, by adding the local graph corresponding to the new user identifier and its associated behavioral label to the existing user portrait graph to update the user portrait graph, the resources and costs required for incremental composition and online training of the user portrait graph can be reduced, thereby improving the data integrity and reliability of the user portrait graph.

[0113] The updated user portrait map is described below with reference to Figure 5. Figure 5 is a schematic diagram of the structure of the updated user portrait map. The dotted box in Figure 5 is the local map corresponding to the user's newly added data, and the area outside the dotted box is the current user portrait map.

[0114] The newly added data in the current cycle includes consumption category b associated with user a, hot posts a and hot topics b, points of interest b in Tieba a associated with user k, and post contents corresponding to tags 6, 7, and 8.

[0115] As shown in Figure 5, the user ID in the current user portrait graph includes user a, but does not include consumption category b associated with user a. Therefore, the behavioral label consumption category b can be added to the local graph and associated with user a in the current user portrait graph. Furthermore, if the user ID in the current user portrait graph does not include user k, a third-order local graph can be constructed based on the hot posts a and hot topics b, interest points b, and post content corresponding to tags 6, tag 7, and tag 8 in Tieba a associated with user k. The interest points b, Tieba a, and the hot posts a and hot topics b corresponding to Tieba a in the local graph are then merged with the same behavioral labels in the current user portrait graph to obtain an updated user portrait graph.

[0116] In this embodiment, by updating the user portrait map based on user incremental data, not only can the timeliness, data integrity and reliability of the user portrait map be enhanced, but also conditions are provided for further improving the quality of conversations.

[0117] FIG6 is a flow chart of a conversation processing method according to another embodiment of the present disclosure.

[0118] As shown in FIG6 , the dialogue processing method includes steps S601 to S604 .

[0119] S601: extract keywords from the historical conversation database to determine a keyword set corresponding to the historical conversation database.

[0120] In the disclosed embodiment, a large language model may be used to extract keywords from each historical conversation in the historical conversation library, and then a keyword set corresponding to the historical conversation library may be constructed based on all the extracted keywords.

[0121] It should be noted that to save storage space in the historical conversation database, the time interval between two adjacent historical conversations can be determined based on their generation time. If the time interval between any two adjacent historical conversations is greater than a preset value (e.g., 2 minutes), the two conversations can be classified as different conversation rounds, thereby dividing all historical conversations into multiple rounds. The large language model is then used to summarize and extract keywords from the conversation content within each round and store them in the historical conversation database.

[0122] S602: Determine the frequency of occurrence of each keyword in the keyword set in each historical conversation and the emotion type of each historical conversation.

[0123] In the embodiment of the present disclosure, the emotional type of historical conversations can be different types determined by dividing the levels according to the degree of emotional fluctuations of the user during the historical chat process. For example, it can be divided into level 1, level 2, level 3, etc. in order from low to high according to the degree of emotional fluctuations. The embodiment of the present disclosure does not limit this.

[0124] In the disclosed embodiment, the frequency of each keyword in each historical conversation can be counted. A large language model can then be used to identify emotional terms in the historical conversations and analyze the user's emotional fluctuations to determine the emotional type of the historical conversations. For example, if a historical conversation contains no emotional terms or contains words such as "bored" that reflect a stable user mood, the emotional type of the historical conversation can be determined as Level 1. Alternatively, if a historical conversation contains words such as "happy," "angry," or a variety of emotional terms, the emotional level of the historical conversation can be determined to be high, and the emotional type can be determined as Level 2 or 3 based on actual business needs.

[0125] S603: Determine the weight of each historical conversation based on the frequency of occurrence of each keyword in each historical conversation and the emotion type of each historical conversation.

[0126] The weight of a historical conversation refers to the importance of each historical conversation in the historical conversation database.

[0127] In the disclosed embodiment, the influence coefficients of the occurrence frequency and emotion type of keywords on the weight of historical conversations can be determined based on experience, and then the product of the occurrence frequency and the corresponding influence coefficient, as well as the product of the emotion type level and the corresponding influence coefficient, are calculated respectively. The values ​​of the two products are added together to obtain the weight of the historical conversation.

[0128] S604: Determine the forgetting coefficient of each historical conversation based on the emotion type of each historical conversation and the time interval between the generation time of the historical conversation and the current time.

[0129] In the disclosed embodiment, the update coefficient for forgetting coefficients of historical conversations generated at different times can be determined based on the time interval between the time the historical conversation was generated and the current time. The update coefficient ranges from 0 to 1. The longer the time interval between the historical conversations, the closer the corresponding update coefficient is to 0. Conversely, the shorter the time interval between the historical conversations, the closer the corresponding update coefficient is to 1.

[0130] In the embodiment of the present disclosure, the calculation formula of the forgetting coefficient of the historical conversation can be shown as the following formula (1):

[0131] f(0)=(e*r)

[0132] Where f(0) represents the forgetting coefficient corresponding to the time when the historical conversation is first stored in the historical conversation database. e represents the level corresponding to the emotion type of the historical conversation, and its value can be 1, 2, or 3. r represents the update coefficient corresponding to the historical conversation.

[0133] It is understandable that as time goes by, users continuously generate historical conversation data with the dialogue system. Therefore, it is necessary to update the historical dialogue library based on the newly added historical conversations. It is also necessary to appropriately delete the conversation content that is older in the historical dialogue library to control the storage cost of the historical dialogue library and improve the data quality in the historical dialogue library. Therefore, the forgetting coefficient of all conversation histories can be updated every time the historical dialogue library is called. The formula for calculating the updated forgetting coefficient can be shown as follows (2):

[0134] f(t)=f(t-1)*(e*r)

[0135] Among them, f(t-1) represents the forgetting coefficient of the historical dialogue after the last update (i.e., the t-1th time), and f(t) represents the forgetting coefficient of the historical dialogue after this and the tth update.

[0136] It should be noted that when the forgetting coefficient of a historical conversation is less than a threshold, the conversation history can be deleted from the historical conversation library. The threshold can be determined based on actual application conditions and is not limited in the present embodiment.

[0137] In this embodiment, the weight and forgetting coefficient of historical conversations are determined based on the frequency of occurrence, emotion type and generation time of keywords in historical conversations, thereby improving the efficiency of calling and managing historical conversation data, thereby controlling the storage cost of the historical conversation library, and providing conditions for improving the efficiency and quality of conversation processing.

[0138] FIG7 is a schematic diagram of the structure of a conversation processing device proposed in an embodiment of the present disclosure.

[0139] As shown in FIG7 , the dialogue processing device 700 includes a first acquisition module 701 , a first determination module 702 , a second acquisition module 703 , a third acquisition module 704 and a generation module 705 .

[0140] The first acquisition module 701 is configured to acquire a topic of a current conversation of a first user and a first historical conversation library associated with the first user.

[0141] The first determining module 702 is configured to determine the similarity between the topic of the current conversation and each historical conversation in the first historical conversation database.

[0142] The second acquisition module 703 is configured to acquire a plurality of candidate historical conversations from the first historical conversation database based on similarity.

[0143] The third acquisition module 704 is configured to acquire a reference dialogue from multiple candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue.

[0144] The generation module 705 is used to generate a reply statement based on the reference conversation and the subject of the current conversation.

[0145] In some embodiments, the third acquisition module 704 is specifically configured to:

[0146] Determine at least one historical conversation with the largest forgetting coefficient and / or the highest weight among the multiple candidate historical conversations;

[0147] A historical conversation having the smallest time interval between the generation time and the current time and smaller than the time threshold in at least one historical conversation is determined as a reference conversation.

[0148] In some embodiments, the third acquisition module 705 is further configured to:

[0149] When the time interval between the generation time of each historical conversation in at least one historical conversation and the current time is greater than a time threshold, a reply statement is generated based on the topic of the current conversation.

[0150] In some embodiments, the generating module 705 is specifically configured to:

[0151] Determine the role of the dialogue system in the current dialogue;

[0152] Generates a response based on the reference conversation, the topic of the current conversation, and prompt information associated with the role type.

[0153] In some embodiments, the generating module 705 is specifically configured to:

[0154] Determine the role type in the current conversation based on the role types used historically by the first user and the chat frequency with each role type; or

[0155] The role type in the current conversation is determined based on the role types historically used by the second user who is similar to the first user and the chat frequency with each role type.

[0156] In some embodiments, the generating module 705 is further configured to:

[0157] Determine a second user who is similar to the first user based on the behavioral tags associated with each user in the user portrait graph.

[0158] In some embodiments, the dialogue processing device 700 further includes a fourth acquisition module, a query module, and an update module.

[0159] The fourth acquisition module is used to obtain user incremental data in the current cycle, wherein the user incremental data includes a user identifier and an associated behavior tag.

[0160] The query module is used to traverse the current user portrait graph based on user identification and behavior tags.

[0161] The updating module is used to update the behavior tags in the current user portrait map based on at least one behavior tag when the current user portrait map contains the user identifier and does not contain at least one behavior tag in the user incremental data, and obtain an updated user portrait map.

[0162] In some embodiments, the update module is further configured to:

[0163] When the current user portrait map does not contain any user identifier in the user incremental data, the current user portrait map is updated based on any user identifier and a behavior tag associated with any user identifier to obtain an updated user portrait map.

[0164] In some embodiments, the dialogue processing device 700 further includes a second determination module, a third determination module, and a fourth determination module.

[0165] The second determination module is used to extract keywords from the historical conversation library and determine the keyword set corresponding to the historical conversation library.

[0166] The third determination module is used to determine the frequency of occurrence of each keyword in the keyword set in each historical conversation and the emotion type of each historical conversation.

[0167] The fourth determination module is used to determine the weight of each historical conversation based on the frequency of occurrence of each keyword in each historical conversation and the emotional type of each historical conversation.

[0168] It should be noted that the above explanation of the dialogue processing method is also applicable to the dialogue processing device of this embodiment and will not be repeated here.

[0169] In this embodiment, the topic of the first user's current conversation and the first user's associated historical conversation database are first obtained. The similarity between the topic of the current conversation and each historical conversation in the first historical conversation database is determined. Based on the similarity, multiple candidate historical conversations are retrieved from the first historical conversation database. Reference conversations are then obtained from the candidate historical conversations based on the forgetting coefficient and weight of each candidate historical conversation. Finally, a reply is generated based on the reference conversations and the topic of the current conversation. Thus, historical conversations that are similar to the topic of the current conversation and highly timely are screened from the user's associated historical conversation database as reference information to assist in the generation of the reply. This improves the reliability and personalization of the reply, enhances the quality of the conversation, and increases user chat satisfaction.

[0170] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, a computer program product, and a computer program.

[0171] According to an embodiment of the present disclosure, the present disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the dialogue processing method described in any of the above embodiments.

[0172] According to an embodiment of the present disclosure, the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the dialog processing method as described in any of the above embodiments.

[0173] According to an embodiment of the present disclosure, the present disclosure provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the dialog processing method as described in any of the above embodiments.

[0174] According to an embodiment of the present disclosure, the present disclosure provides a computer program, which includes computer program code. When the computer program code is run on a computer, the computer executes the dialog processing method as described in any of the above embodiments.

[0175] FIG8 shows a schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0176] As shown in FIG8 , the device 800 includes a computing unit 801, which can calculate the amount of data stored in a read-only memory (ROM).

[0177] The computing unit 801, ROM 802, and RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0178] Various components in device 800 are connected to I / O interface 805, including an input unit 806, such as a keyboard, mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, etc. The communication unit 809 allows device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0179] The computing unit 801 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the conversation processing method. For example, in some embodiments, the conversation processing method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the conversation processing method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the conversation processing method by any other suitable means (e.g., via firmware).

[0180] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0181] The program code for implementing the method of the embodiment of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0182] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer 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 foregoing.

[0183] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0184] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0185] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers, establishing a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0186] It should be noted that the above explanations of the method and device embodiments are also applicable to the electronic device, computer-readable storage medium, computer program product and computer program of the above embodiments, and will not be repeated here.

[0187] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0188] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present disclosure, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. In the description of the present disclosure, the words "if" and "if" used can be interpreted as "at the time of" or "when" or "in response to a determination" or "under the circumstances of".

[0189] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

[0190] All embodiments of the present disclosure may be implemented individually or in combination with other embodiments, and are all considered to be within the scope of protection claimed by the present disclosure.

Claims

1. A conversation processing method, comprising: Obtaining a topic of a current conversation with a first user and a first historical conversation library associated with the first user; Determining the similarity between the topic of the current conversation and each historical conversation in the first historical conversation library; According to the similarity, a plurality of candidate historical conversations are obtained from the first historical conversation library; Based on the forgetting coefficient and weight of each candidate historical conversation, obtaining a reference conversation from the multiple candidate historical conversations; A reply statement is generated based on the reference conversation and the subject of the current conversation.

2. The method according to claim 1, wherein The obtaining of a reference dialogue from the plurality of candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue comprises: Determine at least one historical conversation having the largest forgetting coefficient and / or the highest weight among the multiple candidate historical conversations; A historical conversation in which the time interval between the generation time and the current time is the smallest and smaller than the time threshold among the at least one historical conversation is determined as the reference conversation.

3. The method according to claim 2, wherein: After determining, among the plurality of candidate historical conversations, at least one historical conversation with the largest forgetting coefficient and / or the highest weight, the method further includes: In a case where the time interval between the generation time of each historical conversation in the at least one historical conversation and the current time is greater than the time threshold, the reply statement is generated based on the topic of the current conversation.

4. The method according to any one of claims 1 to 3, wherein The generating of a reply statement based on the reference conversation and the topic of the current conversation includes: Determining the role type of the dialogue system in the current dialogue; The reply statement is generated based on the reference dialogue, the subject of the current dialogue and the prompt information associated with the role type.

5. The method according to claim 4, wherein: Determining the role type of the dialogue system in the current dialogue includes: Determine the role type in the current conversation based on the role types used historically by the first user and the chat frequency with each role type; or The role type in the current conversation is determined based on the role types historically used by the second user who is similar to the first user and the chat frequency with each role type.

6. The method according to claim 5, wherein: Before determining the role type in the current conversation based on the role types used historically by the second user similar to the first user and the chat frequency with each role type, the method further includes: Determine the second user who is similar to the first user based on the behavior tags associated with each user in the user portrait graph.

7. The method according to any one of claims 1 to 6, wherein Also includes: Obtaining incremental user data within the current cycle, wherein the incremental user data includes a user identifier and associated behavior tags; Based on the user identifier and the behavior tag, traverse the current user portrait graph; When the current user portrait map contains the user identifier but does not contain at least one behavior tag in the user incremental data, the behavior tags in the current user portrait map are updated based on the at least one behavior tag to obtain an updated user portrait map.

8. The method of claim 7, wherein: After traversing the current user portrait graph based on the user identifier and the behavior tag, the method further includes: In the case that the current user portrait map does not contain any user identifier in the user incremental data, the current user portrait map is updated based on any user identifier and a behavior tag associated with any user identifier to obtain the updated user portrait map.

9. The method according to any one of claims 1 to 8, wherein The method further comprises: Extracting keywords from the historical conversation database to determine a keyword set corresponding to the historical conversation database; Determining the frequency of occurrence of each keyword in the keyword set in each historical conversation and the sentiment type of each historical conversation; Determining the weight of each historical conversation according to the frequency of occurrence of each keyword in each historical conversation and the emotional type of each historical conversation; The forgetting coefficient of each of the historical conversations is determined according to the emotion type of each of the historical conversations and the time interval between the generation time of the historical conversation and the current time.

10. A conversation processing device, comprising: A first acquisition module is configured to acquire a topic of a current conversation with a first user and a first historical conversation library associated with the first user; a first determining module, configured to determine a similarity between the topic of the current conversation and each historical conversation in the first historical conversation library; A second acquisition module is configured to acquire a plurality of candidate historical conversations from the first historical conversation library based on the similarity; A third acquisition module is configured to acquire a reference dialogue from the plurality of candidate historical dialogues based on the forgetting coefficient and weight of each candidate historical dialogue; A generation module is used to generate a reply statement based on the reference dialogue and the subject of the current dialogue.

11. The device according to claim 10, wherein The third acquisition module is specifically configured to: Determine at least one historical conversation having the largest forgetting coefficient and / or the highest weight among the multiple candidate historical conversations; A historical conversation in which the time interval between the generation time and the current time is the smallest and smaller than the time threshold among the at least one historical conversation is determined as the reference conversation.

12. The device according to claim 11, wherein The third acquisition module is further configured to: In a case where the time interval between the generation time of each historical conversation in the at least one historical conversation and the current time is greater than the time threshold, the reply statement is generated based on the topic of the current conversation.

13. The device according to any one of claims 10 to 12, wherein The generation module is specifically used to: Determining the role type of the dialogue system in the current dialogue; The reply statement is generated based on the reference dialogue, the subject of the current dialogue and the prompt information associated with the role type.

14. The apparatus of claim 13, wherein: The generation module is specifically used to: Determine the role type in the current conversation based on the role types used historically by the first user and the chat frequency with each role type; or The role type in the current conversation is determined based on the role types historically used by the second user who is similar to the first user and the chat frequency with each role type.

15. The apparatus of claim 14, wherein: The generating module is further configured to: Determine a second user who is similar to the first user based on the behavior tags associated with each user in the user portrait graph.

16. The device according to any one of claims 10 to 15, wherein Also includes: A fourth acquisition module is configured to acquire incremental user data within a current period, wherein the incremental user data includes a user identifier and an associated behavior tag; A query module, configured to traverse the current user portrait graph based on the user identifier and the behavior tag; An updating module is used to update the behavior tags in the current user portrait map based on the at least one behavior tag when the current user portrait map contains the user identifier and does not contain at least one behavior tag in the user incremental data, so as to obtain an updated user portrait map.

17. The apparatus of claim 16, wherein: The update module is further configured to: In the case that the current user portrait map does not contain any user identifier in the user incremental data, the current user portrait map is updated based on any user identifier and a behavior tag associated with any user identifier to obtain the updated user portrait map.

18. The device according to any one of claims 10 to 17, wherein The device further comprises: A second determining module is used to extract keywords from the historical conversation library and determine a keyword set corresponding to the historical conversation library; A third determination module is used to determine the frequency of occurrence of each keyword in the keyword set in each historical conversation and the emotion type of each historical conversation; a fourth determining module, configured to determine a weight of each of the historical conversations based on the frequency of occurrence of each keyword in each of the historical conversations and the sentiment type of each of the historical conversations; The fifth determining module is used to determine the forgetting coefficient of each of the historical conversations according to the emotion type of each of the historical conversations and the time interval between the generation time of the historical conversation and the current time.

19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the dialogue processing method according to any one of claims 1 to 9.

20. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: in, The computer instructions are used to cause the computer to execute the dialogue processing method according to any one of claims 1 to 9.

21. A computer program product, characterized in that The computer program comprises a computer program which, when executed by a processor, implements the dialog processing method according to any one of claims 1 to 9.

22. A computer program, characterized in that The computer program includes computer program code, and when the computer program code is run on a computer, the computer is caused to execute the dialog processing method according to any one of claims 1 to 9.

Images