CN117932036A

CN117932036A - Dialogue processing method and device, electronic equipment and storage medium

Info

Publication number: CN117932036A
Application number: CN202410171564.5A
Authority: CN
Inventors: 张思怡; 庞磊; 杨栋; 白云龙
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2024-02-06
Filing date: 2024-02-06
Publication date: 2024-04-26

Abstract

The disclosure provides a dialogue processing method, a dialogue processing device, electronic equipment and a storage medium, relates to the technical field of computers, and particularly relates to the technical field of artificial intelligence such as large language models, natural language processing, knowledge graphs, deep learning and the like. The specific implementation scheme is as follows: acquiring a theme of a current dialogue of a first user and a first historical dialogue library associated with the first user; determining the similarity between the theme of the current dialogue and each historical dialogue in the first historical dialogue library; according to the similarity, a plurality of candidate historical dialogs are obtained from a first historical dialog library; acquiring a reference dialogue from a plurality of candidate historical dialogues based on forgetting coefficients and weights of each candidate historical dialog; a reply sentence is generated based on the topic of the reference dialog, the current dialog.

Description

Dialogue processing method and device, electronic equipment and storage medium

Technical Field

The disclosure relates to the technical field of computers, in particular to the technical field of artificial intelligence such as large language models, natural language processing, knowledge maps, deep learning and the like, and specifically relates to a dialogue processing method, a dialogue processing device, an electronic device and a storage medium.

Background

With the development of artificial intelligence technology, a large language model and application thereof have been widely paid attention to. Therefore, how to improve the individuation degree and accuracy of the reply sentences generated by the large language model in the dialogue becomes a urgent problem to be solved, such as improving the user experience and increasing the application core competitiveness.

Disclosure of Invention

The present disclosure aims to solve, at least to some extent, one of the technical problems in the related art.

Therefore, an object of the present disclosure is to provide a dialogue processing method, apparatus, electronic device, and storage medium, which can improve reliability and individuation degree of generated reply sentences, improve dialogue quality, and increase satisfaction degree of user chat.

According to a first aspect of the present disclosure, there is provided a dialog processing method, including:

acquiring a theme of a current dialogue of a first user and a first historical dialogue library associated with the first user;

determining the similarity between the theme of the current dialogue and each history dialogue in the first history dialogue library;

according to the similarity, a plurality of candidate historical dialogs are obtained from the first historical dialog library;

Acquiring a reference dialogue from the plurality of candidate historical dialogues based on forgetting coefficients and weights of each candidate historical dialog;

a reply sentence is generated based on the reference dialog, the topic of the current dialog.

According to a second aspect of the present disclosure, there is provided a dialog processing device comprising:

the first acquisition module is used for acquiring the theme of the current dialogue of the first user and a first history dialogue library associated with the first user;

The first determining module is used for determining the similarity between the theme of the current dialogue and each history dialogue in the first history dialogue library;

The second acquisition module is used for acquiring a plurality of candidate historical dialogs from the first historical dialog library according to the similarity;

a third obtaining module, configured to obtain a reference dialogue from the plurality of candidate history dialogues based on a forgetting coefficient and a weight of each of the candidate history dialogues;

and the generation module is used for generating a reply sentence based on the reference dialogue and the theme of the current dialogue.

According to a third aspect of the present disclosure, there is provided an electronic device comprising:

At least one processor; and

A memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the dialog processing method as described in the first aspect.

According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the dialog processing method as described in the first aspect.

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

The dialogue processing method, the device, the electronic equipment and the storage medium provided by the disclosure have the following beneficial effects:

In the embodiment of the disclosure, firstly, a subject of a current dialogue of a first user and a first history dialogue library associated with the first user are obtained, similarity between the subject of the current dialogue and each history dialogue in the first history dialogue library is determined, then, a plurality of candidate history dialogues are obtained from the first history dialogue library according to the similarity, a reference dialogue is obtained from the plurality of candidate history dialogues based on forgetting coefficients and weights of each candidate history dialogue, and finally, a reply sentence is generated based on the reference dialogue and the subject of the current dialogue. Thus, by screening the history sentences similar to the current dialogue topic and having high timeliness in the history dialogue library associated with the user, the reply sentences are assisted to be generated as the reference information. Therefore, the reliability and the individuation degree of the reply sentences are improved, the conversation quality is improved, and the satisfaction degree of the chat of the user is increased.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

The foregoing and/or additional aspects and advantages of the present disclosure will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, which serve to better understand the present disclosure, and are not to be construed as limiting the present disclosure, wherein:

FIG. 1 is a flow chart of a dialog processing method according to an embodiment of the present disclosure;

FIG. 2 is a schematic view of a user representation provided by the present disclosure;

FIG. 3 is a flow chart of a dialog processing method according to another embodiment of the present disclosure;

FIG. 4 is a flow chart of a dialog processing method according to another embodiment of the present disclosure;

FIG. 5 is a schematic diagram of an updated user representation map provided by the present disclosure

FIG. 6 is a flow chart of a dialog processing method according to another embodiment of the present disclosure;

FIG. 7 is a schematic diagram of a dialogue processing apparatus according to an embodiment of the disclosure;

fig. 8 illustrates a block diagram of an exemplary electronic device suitable for use in implementing embodiments of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The embodiment of the disclosure relates to the technical fields of artificial intelligence such as large language models, natural language understanding, knowledge maps, deep learning and the like.

Artificial intelligence (ARTIFICIAL INTELLIGENCE), english is abbreviated AI. It is a new technical science for researching, developing theory, method, technology and application system for simulating, extending and expanding human intelligence.

The large language model (Large Language Model, LLM), i.e., the large model, is a deep learning model trained on massive text data. It not only can generate natural language text, but also can deeply understand text meaning, and can process various natural language tasks, such as text abstract, question and answer, translation, etc.

Natural language understanding (Natural Language Understanding, NLU), commonly known as man-machine dialog. Branch disciplines of artificial intelligence. The research uses the electronic computer to simulate the language interaction process of human, so that the computer can understand and use the natural language of human society such as Chinese, english, etc. to realize the natural language communication between human and machine, to replace part of mental labor of human, including inquiring data, solving problem, picking document, assembling data and processing all related natural language information.

The Knowledge map (knowledgegraph), called Knowledge domain visualization or Knowledge domain mapping map in book condition report, is a series of various graphs showing Knowledge development process and structural relationship, and uses visualization technology to describe Knowledge resources and their carriers, and excavate, analyze, construct, draw and display Knowledge and their interrelationships.

Deep learning is the inherent regularity and presentation hierarchy of learning sample data, and the information obtained during such learning is helpful in interpreting data such as text, images and sounds. The final goal of deep learning is to enable a machine to analyze learning capabilities like a person, and to recognize text, images, and sound data.

In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information of the user accord with the regulations of related laws and regulations, and the public order colloquial is not violated.

The following describes a dialogue processing method, a device, an electronic apparatus, and a storage medium according to an embodiment of the present disclosure with reference to the accompanying drawings.

It should be noted that, the main execution body of the session processing method in this embodiment is a session processing apparatus, and the apparatus may be implemented in software and/or hardware, and the apparatus may be configured in an electronic device, and the electronic device may include, but is not limited to, a terminal, a server, and the like.

In the present disclosure, the dialog processing device may be configured in a dialog system of any application or website, etc., and the dialog system may generate reply content to perform a dialog with a user by the dialog processing method provided in the embodiments of the present disclosure.

Fig. 1 is a flow chart illustrating a dialog processing method according to an embodiment of the present disclosure.

As shown in fig. 1, the dialogue processing method includes:

S101: and acquiring a theme of the current dialogue of the first user and a first historical dialogue library associated with the first user.

The first user may be any user that triggers a dialogue system. Triggering the operation of the dialog system may include clicking a chat interface entering the dialog system, clicking a specific control in a program or website with a dialog function, etc., which may be preset according to actual application needs, and the disclosure is not limited thereto.

In the embodiment of the disclosure, when the current dialogue interface includes dialogue content of the chat between the first user and the system, the dialogue content in the interface may be analyzed and summarized to obtain the topic of the current dialogue.

It should be noted that, in this disclosure, the current session may be initiated not only by the first user, but also by the session system actively. That is, when the first user triggers the dialogue system, but no dialogue content exists in the current dialogue interface, the dialogue system may analyze the behavior habit of the first user according to the pre-constructed user portrait graph, determine the topic that the first user is most likely to be interested in at present, as the topic of the current dialogue, and initiate the dialogue based on the topic.

How the user profile is constructed will be described with reference to fig. 2, and fig. 2 is a schematic diagram of the structure of the user profile.

In the present disclosure, behavior data of individual users of programs or websites, etc. to which the dialog system is applied may be collected by an offline manner, etc., and behavior tags associated with each user, such as a bar name, a point of interest, a consumption class, a post tag (tag), etc., may be extracted from the behavior data. For example, after behavior habits of a user a and a user b are collected and behavior labels are extracted for the behavior habits of the user a and the user b respectively, the behavior labels associated with the user a can be obtained and include a point of interest a, a point of interest b, a consumption class b and post contents corresponding to tag1 and tag2, and the behavior labels associated with the user b include hot posts a and hot topics b, the point of interest b, the consumption class b and post contents corresponding to tag3, tag4 and tag5 in a bar a.

Then, the user identifications a and b and the associated behavior labels are used as nodes in the user portrait graph, and connection relations between the user identification nodes and the behavior label nodes are determined according to clicking, consumption and other behaviors of the user, so that the user portrait graph is constructed, and the user portrait graph is shown in figure 2. As can be seen from fig. 2, the vector representation of the user identifier node for each behavior label node can be obtained based on the number of clicks of the user on the interest point, the number of consumption of the user classification, the browsing duration of the post content or the bar, and the weight determined by the behavior such as praise.

In the embodiment of the disclosure, the similarity of the vectors between the first user and the behavior label can be calculated according to the representation of each node vector in the user portrait graph. A plurality of candidate topics is then determined based on the order in which the similarity is greater than the similarity. And then sorting and screening the multiple candidate topics according to the latest hot topics acquired by the ways such as bar pasting and the latest behavior habits of the user, and determining a candidate topic with the highest similarity as the topic which is most likely to be interested by the first user at present.

In the embodiment of the disclosure, the history dialogue library associated with each user may be pre-constructed according to the history dialogue content associated with the user. When the conversation is conducted, a first historical conversation library associated with the first user can be found according to the identification (such as a user name, an account number and the like) of the first user.

S102: and determining the similarity between the theme of the current dialogue and each historical dialogue in the first historical dialogue library.

It should be noted that, the first historical dialog library may include abstracts, keywords, and the like corresponding to each historical dialog. Therefore, when calculating the similarity between the topic of the current dialogue and each history dialogue in the first history dialogue library, the similarity may be between the topic of the current dialogue and the keyword of each history dialogue, or may be between the topic of the current dialogue and the abstract of each history dialogue, or the like, which is not limited in this disclosure.

S103: and acquiring a plurality of candidate historical dialogs from the first historical dialog library according to the similarity.

In the embodiment of the disclosure, the first N corresponding history dialogs may be determined as candidate history dialogs according to the order of the similarity from high to low. Wherein N may be any positive integer, such as 5, 10, etc., which is not limited in this disclosure.

S104: a reference dialog is obtained from the plurality of candidate historical dialogs based on the forgetting coefficient and the weight of each candidate historical dialog.

The forgetting coefficient, which indicates the influence degree of each history dialogue on the current reply content, can be determined according to the emotion fluctuation degree of the user in the history dialogue and the time interval between the corresponding generation time and the current time. The larger the mood fluctuation of the user is, the smaller the time interval is, the larger the forgetting coefficient is, and the larger the influence degree on the current reply content is.

The forgetting coefficient is updated with time, and the longer the interval between the history dialogue generation time and the current time is, the smaller the forgetting coefficient is. When the forgetting coefficient of the history dialogue is smaller than a certain value, the history dialogue can be deleted in the first history dialogue library.

The weight of the candidate history dialogs may be a value determined according to the occurrence frequency of keywords representing dialog intention in the history dialogs, the emotion fluctuation degree of the user at the time of dialogs, and the like, and is used for describing the importance degree of each history dialogs in the history dialogs library.

It should be noted that, when the history dialogue is stored in the first history repository, the weight and the forgetting rate of the history dialogue may be determined and stored in association with the history dialogue. Therefore, when the reference dialogue is selected, the forgetting coefficient and the weight of each candidate historical dialogue can be directly obtained from the first historical dialogue library.

Alternatively, at least one history dialog having the highest forgetting coefficient and/or highest weight among the plurality of candidate history dialogues may be first determined. And then determining one of the at least one historical dialogs, which has the smallest time interval between the generation time and the current time and is smaller than the time threshold value, as the reference dialog.

It should be noted that, when the time interval between the generation time and the current time of the historical dialog is smaller than the time threshold, the timeliness of the historical dialog is high, and the historical dialog may have a better reference value for generating the reply sentence. The time threshold may be a value determined according to a service requirement, for example, may be 10 days, 30 days, or the like, which is not limited in the present disclosure.

In the embodiment of the disclosure, the reference dialogue for assisting in generating the reply sentence is determined based on the forgetting coefficient, the weight and the generation time of the history dialogue, so that the screened reference dialogue can be ensured to be more reliable and have more reference value, and the quality and the individuation degree of the reply sentence are improved.

S105: a reply sentence is generated based on the topic of the reference dialog, the current dialog.

In the embodiment of the disclosure, the reply sentence can be generated by inputting the topics of the reference dialog and the current dialog into the language model of the dialog system.

It should be noted that the language model may be a large language model (Large Language Model, LLM), such as GPT, a discourse, etc., or may be other language models capable of generating reply sentences, which is not limited in this disclosure.

Alternatively, in the case where the time interval between the generation time of each of the at least one history dialogs and the current time is greater than the time threshold, the reply sentence may be generated based on the subject of the current dialog.

It will be appreciated that when the time of generation of the historical dialog is greater than the time threshold, the historical dialog may not accurately reflect the user's current dialog preference, possibly resulting in bias and semantic errors in the reply sentence. So that when generating the reply sentence, the reply sentence related to the current dialogue topic can be directly generated without referring to the history dialogue information. Therefore, the history dialogue with long production time can be prevented from being used as the reference information for sentence generation, and the accuracy and timeliness of reply sentences are further ensured.

In this embodiment, a subject of a current dialogue of a first user and a first history dialogue library associated with the first user are firstly obtained, a similarity between the subject of the current dialogue and each history dialogue in the first history dialogue library is determined, then a plurality of candidate history dialogues are obtained from the first history dialogue library according to the similarity, a reference dialogue is obtained from the plurality of candidate history dialogues based on forgetting coefficients and weights of each candidate history dialogue, and finally a reply sentence is generated based on the reference dialogue and the subject of the current dialogue. Thus, by screening the history sentences similar to the current dialogue topic and having high timeliness in the history dialogue library associated with the user, the reply sentences are assisted to be generated as the reference information. Therefore, the reliability and the individuation degree of the reply sentences are improved, the conversation quality is improved, and the satisfaction degree of the chat of the user is increased.

Fig. 3 is a flow chart illustrating a dialogue processing method according to another embodiment of the disclosure.

As shown in fig. 3, the dialogue processing method includes:

S301: and acquiring a theme of the current dialogue of the first user and a first historical dialogue library associated with the first user.

S302: and determining the similarity between the theme of the current dialogue and each historical dialogue in the first historical dialogue library.

S303: and acquiring a plurality of candidate historical dialogs from the first historical dialog library according to the similarity.

S304: a reference dialog is obtained from the plurality of candidate historical dialogs based on the forgetting coefficient and the weight of each candidate historical dialog.

The descriptions of S301 to S304 may be specifically referred to the above embodiments, and are not repeated herein.

It should be noted that in some application scenarios, such as games, text creation, etc., a dialogue system may be required to simulate the speaking characteristics of a character to perform a dialogue with a user. Therefore, before generating the reply sentence, it is also necessary to determine the role of the target simulation.

S305: the type of role of the dialog system in the current dialog is determined.

Wherein the character types may be classified according to at least one of gender, age range, occupation, speaking mood or habit, character relationship, and the like. Or the characters in the dialog system may also include certain specific characters set by movies, games, etc., which are not limited by the present disclosure.

It should be noted that the type of role played by the dialog system during a round of dialog is fixed and well-defined before each round of dialog begins. The type of role played by the dialog system during a round of dialog may be selected by the first user according to his own needs or may be selected by the dialog system according to the historical preferences associated with the first user. After the selected role type to play, it is stored in the system, and then the stored role type is called each time a reply sentence is generated.

Optionally, if the first user does not select a role type, the dialog system may determine, according to the role type used by the first user historically and the chat frequency with each role type, the role type with the highest chat frequency as the role type in the current dialog.

Or in the case that the first user does not have a character type used in history, the character type in the current dialogue may be determined according to the character type used in history by the second user similar to the first user and the chat frequency with each character type.

In the embodiment of the disclosure, the character type with highest historical use frequency is determined as the target character type of the current dialogue, so that the preference and the demand of the user can be better met, the time cost on selecting the characters is reduced, and the satisfaction degree of the user is improved.

Alternatively, a second user similar to the first user may be determined based on the behavior tags associated with each user in the user profile.

It will be appreciated that if different users have similar behavioural habits, they are likely to like the same type of character. Therefore, the candidate subject corresponding to the first user can be determined according to the behavior label associated with the first user in the user portrait atlas, then the behavior labels of other users in the atlas are checked, the behavior labels are matched with the candidate subject, and the user with the highest similarity is determined to be the second user similar to the first user.

Therefore, through the behavior labels associated with each user in the user portrait atlas, the second user similar to the first user is determined, the accuracy and reliability of the second user determination can be improved, and conditions are provided for improving the accuracy of determining the character type preferred by the first user.

It should be noted that, the type of the character used recently by the first user or the second user may also be determined as the type of the character in the current dialog of the dialog system, which is not limited in this disclosure.

S306: based on the reference dialog, the topic of the current dialog and the prompt information associated with the character type, a reply sentence is generated.

The prompt information associated with the role type can include description information corresponding to the sex, the age range, the occupation, the speaking mood or habit, the relationship of the player characters and the like corresponding to the role type. For example, the hint information may be "middle-aged female, teacher, severity", or "male, father, and familiarity", etc.

In this embodiment, after determining the reference dialogue generated by the sentence, the role type of the dialogue system in the current dialogue is determined, and then the reply sentence is generated based on the reference dialogue, the topic of the current dialogue and the prompt information associated with the role type. Therefore, the reply sentences with the character individuation styles are generated based on the prompt sentences corresponding to the target character types, so that the diversity and the quality of the reply sentences can be further improved, and the interaction of users is enhanced.

Fig. 4 is a flow chart illustrating a dialogue processing method according to another embodiment of the disclosure.

As shown in fig. 4, the dialogue processing method includes:

S401: user increment data in the current period is obtained.

The user increment data comprises user identifications and associated behavior labels.

It will be appreciated that the user profile is constructed based on historical user behavior data, but new user behavior data is continually generated and the degree of influence of the historical data on analysis of the user's preferences is reduced over time, so that the effectiveness of previously constructed user profiles in determining the dialog topic preferences of the target user may be reduced. Therefore, the user portrait map needs to be updated regularly to ensure that the currently used user portrait map is the latest and the most complete, thereby ensuring the accuracy of the dialog theme determined by the dialog system.

In the embodiment of the disclosure, an update period may be preset, and a behavior label may be extracted from the user behavior data newly generated in the current period, and associated with the user identifier to be used as user increment data to update the user portrait map.

It should be noted that the update period may be determined according to the actual situation. For example, in the case where the magnitude of the newly added user behavior data is large, the period may be shortened, which is not limited by the present disclosure.

S402: traversing the current user portrait map based on the user identification and the behavior label.

In the embodiment of the disclosure, whether the user identifier is included or not may be searched in the current user portrait map sequentially according to each user identifier in the user increment data. And then, determining whether the behavior label associated with the user identifier in the current user portrait atlas is the same as the behavior label associated with the user identifier according to the behavior label associated with the user identifier in the user increment data.

S403: and under the condition that the current user portrait map contains the user identifier and does not contain at least one behavior label in the user increment data, updating the behavior label in the current user portrait map based on the at least one behavior label, and acquiring the updated user portrait map.

In the embodiment of the disclosure, when the user portrait map contains any user identifier in the user incremental data, whether the behavior label associated with the user identifier corresponds to the behavior label associated with the user identifier in the user incremental data can be searched in the current user portrait map. Under the condition that at least one behavior label in the user increment data does not exist in the current user portrait graph, the behavior label which is not contained can be added into the current user portrait graph as a neighbor node of the user identifier, a connection relation is established between the neighbor node and the user identifier, then an aggregation function is adopted to train newly added nodes and edges in the user portrait graph, and updating of the user portrait graph is completed.

Optionally, under the condition that the current user portrait map does not contain any user identifier in the user increment data, updating the current user portrait map based on any user identifier and a behavior label associated with any user identifier, and obtaining the updated user portrait map.

In the embodiment of the disclosure, when the user portrait map does not contain any user identifier in the user incremental data, a local map can be constructed based on any user identifier and a behavior label associated with any user identifier, and then the local map and the current user portrait map are combined by adopting an aggregation function to complete updating of the user portrait map. Therefore, the user portrait map is updated by adding the new user identifier and the local map corresponding to the associated behavior label in the existing user portrait map, so that the resources and cost required by incremental composition and online training of the user portrait map can be reduced, and the data integrity and reliability of the user portrait map are improved.

The updating of the user profile is described below in connection with fig. 5. FIG. 5 is a schematic diagram of the structure of the updated user portrait map, and the partial map corresponding to the newly added data of the user is shown in the dashed line box in FIG. 5, and the current user portrait map is shown outside the dashed line box. The newly added data in the current period comprises a consumption class b associated with a user a, hot posts a and hot topics b in a post bar a associated with a user k, interest points b and post contents corresponding to tags 6, 7 and 8.

As shown in FIG. 5, the user identification in the current user profile contains user a but does not contain the consumption class b associated with user a, so that behavior tag consumption class b can be added to the local profile to be associated with user a in the current user profile. And, if the user identifier in the current user portrait map does not include the user k, a 3-order local map is constructed according to the hot label a and the hot topic b in the bar a, the interest point b and the post contents corresponding to the tag6, the tag7 and the tag8 associated with the user k, and the interest point b, the bar a and the hot label a and the hot topic b corresponding to the bar a in the local map are combined with the same behavior label in the current user portrait map, so that the updated user portrait map is obtained.

In this embodiment, the user portrait map is updated by being based on user delta data. Not only can the timeliness, the data integrity and the reliability of the user portrait map be enhanced, but also conditions are provided for further improving the conversation quality.

Fig. 6 is a flow chart illustrating a dialogue processing method according to another embodiment of the disclosure.

As shown in fig. 6, the dialogue processing method includes:

s601: and extracting keywords from the history dialogue library, and determining a keyword set corresponding to the history dialogue library.

In the disclosure, keyword extraction can be performed on each history dialogue in the history dialogue library by using a large language model, and then a keyword set corresponding to the history dialogue library is constructed according to all extracted keywords.

In order to save the storage space required by the history dialogue library, when the interval between the generation times of two history dialogues adjacent to each other in any generation order is greater than a preset value (such as 2 minutes), the two history dialogues may be determined as different dialogue rounds, so that all the history dialogues may be divided into dialogue contents in a plurality of rounds. And then summarizing the dialogue content in each turn and extracting key words by using a large language model, and storing the abstract summary and the key words into a history dialogue library.

S602: the frequency of occurrence of each keyword in each historical dialog in the keyword set and the emotion type of each historical dialog are determined.

In the present disclosure, emotion types of a history dialogue may be different types determined by classifying according to emotion fluctuation degrees of users in a history chat process, for example, the emotion types may be classified into 1 level, 2 level, 3 level, etc. according to the order of low emotion fluctuation degrees, which is not limited in the present disclosure.

In the embodiment of the disclosure, the occurrence frequency of each keyword in each historical dialog can be counted, and the emotion type of the historical dialog is determined by utilizing a large language model to identify words containing emotion in the historical dialog and analyzing the emotion fluctuation degree of the user. For example, in the case where words reflecting emotion are not included in the history dialogue or words reflecting user emotion stabilization such as "boring" are included, it may be determined that the emotion type of the history dialogue is level 1; or when the history dialogue contains words of 'happy', 'angry', or a plurality of types of description emotion, the emotion level of the history dialogue can be identified to be higher, and the emotion type can be determined to be level 2 or level 3 according to the actual service requirement.

S603: and determining the weight of each historical dialog according to the occurrence frequency of each keyword in each historical dialog and the emotion type of each historical dialog.

The weight of the history dialogue refers to the importance degree of each history dialogue in the history dialogue library.

In the embodiment of the disclosure, the influence coefficients of the occurrence frequency and the emotion type of the keyword on the weight of the historical dialog can be determined according to experience, then the products of the occurrence frequency and the corresponding influence coefficients and the products of the emotion type level and the corresponding influence coefficients are calculated respectively, and the values of the two products are added to obtain the weight of the historical dialog.

S604: and determining the forgetting coefficient of each historical dialogue according to the emotion type of each historical dialogue and the time interval between the generation time and the current time of the historical dialogue.

In the embodiment of the disclosure, the update coefficient of the historical dialog generated at different time when the forgetting coefficient is updated can be determined according to the time interval between the generation time and the current time of the historical dialog. The larger the time interval corresponding to the history dialogue is, the more the corresponding update coefficient is approaching 0, whereas the smaller the time interval corresponding to the history dialogue is, the more the corresponding update coefficient is approaching 1.

In the present disclosure, the calculation formula of the forgetting coefficient of the history dialogue may be shown as the following formula (1):

f(0)＝(e*r)

Wherein f (0) represents a forgetting coefficient corresponding to the history dialogue when the history dialogue is just stored in the history dialogue library. e represents a level corresponding to the emotion type of the history dialogue, and the value of e may be 1, 2 or 3. r represents the update coefficient corresponding to the history dialogue.

It can be understood that, as time increases, the user generates historical dialogue data with the dialogue system continuously, so that the historical dialogue library needs to be updated based on the newly added historical dialogue, and dialogue content with far time in the historical dialogue library needs to be deleted appropriately to control the storage cost of the historical dialogue library and improve the data quality in the historical dialogue library. Therefore, each time the history dialogue library is called, the forgetting coefficients of all the dialogue histories can be updated, and the formula for calculating the updated forgetting coefficients can be shown as the following formula (2):

f(t)＝f(t-1)*(e*r)

wherein f (t-1) represents the forgetting coefficient of the history dialogue after the last (i.e. the t-1 th) update, and f (t) represents the forgetting coefficient of the history dialogue after the present and the t-th updates.

It should be noted that, when the forgetting coefficient of the history dialog is smaller than the threshold value, the dialog history may be deleted in the history dialog library. The threshold may be determined according to the actual application, which is not limited in this disclosure.

In this embodiment, the weight and forgetting coefficient of the history dialogue are determined based on the occurrence frequency, emotion type and generation time of the keywords in the history dialogue, so that the efficiency of calling and managing the history dialogue data is improved, the storage cost of the history dialogue library can be controlled, and conditions are provided for improving the efficiency and quality of dialogue processing.

Fig. 7 is a schematic structural diagram of a dialogue processing device according to an embodiment of the disclosure.

As shown in fig. 7, the session processing apparatus 700 includes:

A first obtaining module 701, configured to obtain a subject of a current session of a first user and a first historical session library associated with the first user;

A first determining module 702, configured to determine a similarity between a topic of the current session and each historical session in the first historical session library;

a second obtaining module 703, configured to obtain a plurality of candidate history dialogs from the first history dialog library according to the similarity;

A third obtaining module 704, configured to obtain a reference dialogue from a plurality of candidate history dialogues based on forgetting coefficients and weights of each candidate history dialog;

A generating module 705, configured to generate a reply sentence based on the reference dialog and the topic of the current dialog.

In some embodiments, the third obtaining module 704 is specifically configured to:

determining at least one history dialogue with the largest forgetting coefficient and/or highest weight in the plurality of candidate history dialogues;

And determining one of the at least one historical dialogs, which has the smallest time interval between the generation time and the current time and is smaller than the time threshold, as the reference dialog.

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

And generating a reply sentence based on the topic of the current dialogue under the condition that the time interval between the generation time of each historical dialogue in at least one historical dialogue and the current time is larger than a time threshold value.

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

Determining the role type of the dialogue system in the current dialogue;

Based on the reference dialog, the topic of the current dialog and the prompt information associated with the character type, a reply sentence is generated.

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

Determining the role type in the current dialogue according to the role type used by the first user history and the chat frequency of each role type; or alternatively

The character type in the current conversation is determined according to the character type used by the second user similar to the first user and the chat frequency of each character type.

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

A second user similar to the first user is determined based on the behavior tags associated with each user in the user profile.

In some embodiments, the dialog processing device 700 further includes:

a fourth acquisition module, configured to acquire user incremental data in a current period, where the user incremental data includes a user identifier and an associated behavior tag;

The query module is used for traversing the current user portrait atlas based on the user identification and the behavior label;

and the updating module is used for updating the behavior label in the current user portrait graph based on at least one behavior label under the condition that the current user portrait graph contains the user identification and does not contain at least one behavior label in the user increment data, and acquiring the updated user portrait graph.

In some embodiments, the update module is further to:

and under the condition that the current user portrait map does not contain any user identifier in the user increment data, updating the current user portrait map based on any user identifier and a behavior label associated with the user identifier, and acquiring the updated user portrait map.

In some embodiments, the dialog processing device 700 further includes:

the second determining module is used for extracting keywords from the history dialogue library and determining a keyword set corresponding to the history dialogue library;

The third determining module is used for determining the occurrence frequency of each keyword in each historical dialogue in the keyword set and the emotion type of each historical dialogue;

A fourth determining module, configured to determine a weight of each historical dialog according to an occurrence frequency of each keyword in each historical dialog and an emotion type of each historical dialog;

It should be noted that the foregoing explanation of the session processing method is also applicable to the session processing apparatus of the present embodiment, and will not be repeated here.

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

Fig. 8 illustrates a schematic block diagram of an example electronic device 800 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 8, the apparatus 800 includes a computing unit 801 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 802 or a computer program loaded from a storage unit 808 into a Random Access Memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to the bus 804.

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, and the like; 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, or the like. The communication unit 809 allows the device 800 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The computing unit 801 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of 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 respective methods and processes described above, such as a dialogue processing method. For example, in some embodiments, the dialog processing method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 800 via ROM 802 and/or communication unit 809. When a computer program is loaded into RAM 803 and executed by computing unit 801, one or more steps of the dialog processing method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the dialog processing method by any other suitable means (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here 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 a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, 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: local Area Networks (LANs), wide Area Networks (WANs), the internet, and blockchain networks.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical hosts and VPS service ("Virtual PRIVATE SERVER" or simply "VPS") are overcome. The server may also be a server of a distributed system or a server that incorporates a blockchain.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present disclosure, the meaning of "a plurality" is at least two, such as two, three, etc., unless explicitly specified otherwise. In the description of the present disclosure, the words "if" and "if" as used may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in the … … case".

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. A dialog processing method, comprising:

2. The method of claim 1, wherein the obtaining a reference dialogue from the plurality of candidate historical dialogues based on forgetting coefficients and weights for each of the candidate historical dialogues comprises:

Determining at least one history dialogue with the largest corresponding forgetting coefficient and/or highest weight in the plurality of candidate history dialogues;

and determining one historical dialog with the minimum time interval between the generation time and the current time in the at least one historical dialog and less than a time threshold as the reference dialog.

3. The method of claim 2, wherein in said determining said plurality of candidate historical dialogs, at least one historical dialog having a highest forgetting coefficient and/or highest weight, further comprising, after:

And generating a reply sentence based on the topic of the current dialogue under the condition that the time interval between the generation time of each historical dialogue and the current time in the at least one historical dialogue is larger than a time threshold value.

4. The method of claim 1, wherein the generating a reply sentence based on the reference dialog, the topic of the current dialog, comprises:

Determining the role type of a dialogue system in the current dialogue;

And generating the reply sentence based on the topic of the current dialogue and the prompt information associated with the role type.

5. The method of claim 4, wherein the determining the type of role of the dialog system in the current dialog comprises:

And determining the character type in the current dialogue according to the character type used by a second user similar to the first user in history and the chat frequency of each character type.

6. The method of claim 5, wherein prior to determining the character type in the current conversation based on the character type historically used by a second user similar to the first user and the chat frequency with each character type, further comprising:

And determining a second user similar to the first user according to the behavior label associated with each user in the user portrait graph.

7. The method of claim 6, further comprising:

acquiring user increment data in a current period, wherein the user increment data comprises a user identifier and an associated behavior label;

Traversing the current user portrait atlas based on the user identifier and the behavior label;

And under the condition that the current user portrait map contains the user identifier and does not contain at least one behavior label in the user increment data, updating the behavior label in the current user portrait map based on the at least one behavior label, and acquiring the updated user portrait map.

8. The method of claim 7, wherein after traversing the current user representation based on the user identification and the behavior tag, further comprising:

9. The method of any one of claims 1-8, wherein the method further comprises:

extracting keywords from the history dialogue library, and determining a keyword set corresponding to the history dialogue library;

Determining the occurrence frequency of each keyword in each historical dialogue in the keyword set and the emotion type of each historical dialogue;

determining the weight of each historical dialogue according to the occurrence frequency of each keyword in each historical dialogue and the emotion type of each historical dialogue;

and determining the forgetting coefficient of each historical dialogue according to the emotion type of each historical dialogue and the time interval between the generation time and the current time of the historical dialogue.

10. A dialog processing device comprising:

11. The apparatus of claim 10, wherein the third acquisition module is specifically configured to:

12. The apparatus of claim 11, wherein the third acquisition module is further configured to:

13. The apparatus of claim 10, wherein the generating module is specifically configured to:

Determining the role type of a dialogue system in the current dialogue;

14. The apparatus of claim 13, wherein the generating module is specifically configured to:

15. The apparatus of claim 14, wherein the generation module is further configured to:

16. The apparatus of claim 15, further comprising:

A fourth obtaining module, configured to obtain user incremental data in a current period, where the user incremental data includes a user identifier and an associated behavior tag;

the query module is used for traversing the current user portrait atlas based on the user identifier and the behavior label;

And the updating module is used for updating the behavior label in the current user portrait graph based on at least one behavior label under the condition that the current user portrait graph contains the user identifier and does not contain at least one behavior label in the user increment data, and acquiring the updated user portrait graph.

17. The apparatus of claim 16, wherein the update module is further to:

18. The apparatus of any of claims 1-8, wherein the apparatus further comprises:

a third determining module, configured to determine a frequency of occurrence of each keyword in each historical dialog in the keyword set, and an emotion type of each historical dialog;

and a fifth determining module, configured to determine a forgetting coefficient of each historical dialog according to an emotion type of each historical dialog and a time interval between a generation time and a current time of the historical dialog.

19. An electronic device, comprising:

At least one processor; and

A memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the dialog processing method of any of claims 1-9.

20. A non-transitory computer readable storage medium storing computer instructions, wherein the computer instructions are for causing the computer to perform the dialog processing method of any of claims 1-9.

21. A computer program product comprising a computer program which, when executed by a processor, implements the dialog processing method of any of claims 1-9.