Session knowledge management agent auxiliary method and device based on large model and electronic equipment

Abstract

The invention discloses a session knowledge management agent assisting method and device based on a large model and electronic equipment. According to the embodiment of the invention, word segmentation is carried out on each session message contained in the group message to calculate the first correlation between words so as to determine at least one subject word in the group message; meanwhile, calculating a second correlation between each session message and at least one session message adjacent in the time sequence to determine an associated message of each session message so as to form a theme session group aiming at the specific theme; calculating a second correlation of each word vector of a plurality of session messages contained in the subject term to determine a first correlation word of each subject term, calculating a third correlation of the first correlation word and a word adjacent in position to determine a second correlation word associated with the third correlation word, and finally generating a subject sequence for the corresponding subject term. The efficiency and the accuracy of group message management can be greatly improved, and group members can quickly acquire key information and know backlogs.

Description

Technical Field

[0001] The present application relates to the field of text processing technology, and in particular to a conversational knowledge management agent assistance method and device and electronic equipment based on a large model. Background Art

[0002] In today's era of digital information explosion, social networking services (SNS) and various online communication tools are widely used for personal and business communication. Social and work groups, in particular, are flooded with messages and their content is complex. Group managers often spend considerable time sifting through these messages to identify key information, such as to-do items, important notifications, or decision-making rationales. However, due to this information overload, manual sifting is inefficient and can easily miss important content.

[0003] Existing message management technologies have limitations when handling such complex scenarios, making it difficult to meet the requirements for efficient and accurate knowledge management and information aggregation within group messages. For example, traditional keyword search methods fail to understand contextual semantics, resulting in inaccurate extraction results. Furthermore, information in group messages is often unstructured, containing a large amount of fragmented content, redundant information, and even noise, further complicating information extraction.

[0004] Therefore, there is an urgent need for an intelligent management solution for group messages based on a large model, which can use natural language processing (NLP) technology and deep learning models to extract meaningful information from massive unstructured messages. Summary of the Invention

[0005] The embodiments of the present application provide a conversation knowledge management agent assistance method and device and an electronic device based on a large model to solve the defect of low efficiency in managing group messages based only on keywords in the prior art.

[0006] To achieve the above technical objectives, the present application proposes a large-model-based conversational knowledge management agent-assisted method, including:

[0007] Obtaining a first group message, wherein the first group message includes a plurality of conversation messages sent by a plurality of group members;

[0008] Performing word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message;

[0009] Calculating a first correlation between the words based on the word vectors, position information of each corresponding word in the corresponding conversation message, and a time sequence of the corresponding conversation messages in the first group message;

[0010] determining, based on the first correlation, at least one topic word in the first group message;

[0011] Calculating a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence;

[0012] Determining, based on the second correlation, associated messages of each conversation message containing the keyword to form at least one topic conversation group, wherein each topic conversation group corresponds to a topic word and includes a plurality of conversation messages that are continuous or discontinuous in time sequence;

[0013] For each topic conversation group in the at least one topic conversation group, calculating a second correlation between word vectors of the plurality of conversation messages contained therein, wherein the second correlation represents a semantic correlation between a word vector in a conversation message and word vectors in other conversation messages in the topic conversation group;

[0014] Determining a first related word corresponding to each subject word based on the second relatedness;

[0015] Calculating a third correlation between the first related word and words adjacent in position in the respective conversation messages;

[0016] determining, based on the third correlation, a second related word associated with the first related word;

[0017] A topic sequence for corresponding topic words is generated based on the first related words and the second related words.

[0018] According to an embodiment of the present application, a conversation message includes sender information, receiver information, sending time information, and sending content.

[0019] According to an embodiment of the present application, obtaining the first group message includes: determining the data type of each conversation message in the obtained first group message; and converting it into text data using a corresponding machine learning model according to the data type.

[0020] According to an embodiment of the present application, converting the data type into text data using a corresponding machine learning model includes: when it is determined that the conversation message belongs to an image data type, using an image recognition model to extract at least one image feature of the conversation message; determining at least one object contained in the conversation message based on the image feature; determining a text description corresponding to the conversation message based on object information of the object; and using the text description as the text data.

[0021] According to an embodiment of the present application, converting the data type into text data using a corresponding machine learning model includes: when it is determined that the conversation message belongs to a voice data type, converting the conversation message into text using a voice recognition model; and using the text as the text data.

[0022] According to an embodiment of the present application, the conversational knowledge management agent-assisted method based on the large model further includes:

[0023] For each word in each conversation message in the topic conversation group, a word vector is calculated, which includes a first result containing a position feature of the topic word corresponding to the topic conversation group in the conversation message and a second result containing a context feature of the topic word in the topic conversation group;

[0024] Performing feature fusion calculation on the first result and the second result to generate a fusion feature;

[0025] Calculating the opinion tendency of the conversation message regarding the keyword based on the fusion feature; and

[0026] The topic sequence corresponding to the topic conversation group is updated according to the opinion tendency.

[0027] The present application also provides a large-model-based conversational knowledge management agent auxiliary device, including:

[0028] An acquisition module, configured to acquire a first group message, wherein the first group message includes a plurality of conversation messages sent by a plurality of group members;

[0029] a preprocessing module, configured to perform word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message;

[0030] a first calculation module, configured to calculate a first correlation between the words based on the word vectors, position information of each word in the corresponding conversation message, and a time sequence of the corresponding conversation messages in the first group message;

[0031] a first determining module, configured to determine at least one keyword in the first group message based on the first correlation;

[0032] a second calculation module, configured to calculate a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence;

[0033] a second determining module configured to determine, based on the second correlation, associated messages of each conversation message containing the keyword to form at least one topic conversation group, wherein each topic conversation group corresponds to a topic word and includes a plurality of conversation messages that are continuous or discontinuous in chronological order;

[0034] a third calculation module, configured to calculate, for each topic conversation group in the at least one topic conversation group, a second correlation between word vectors of the plurality of conversation messages contained therein, wherein the second correlation represents a semantic correlation between a word vector in a conversation message and word vectors in other conversation messages in the topic conversation group;

[0035] a third determining module, configured to determine a first related word corresponding to each subject word according to the second relevance;

[0036] a fourth calculation module, configured to calculate a third correlation between the first related words and words adjacent to them in respective conversation messages;

[0037] a fourth determining module, configured to determine a second related word associated with the first related word according to the third correlation;

[0038] A generating module is used to generate a topic sequence for the corresponding topic word based on the first related word and the second related word.

[0039] An embodiment of the present application further provides an electronic device, including:

[0040] Memory, used to store programs;

[0041] A processor is used to run the program stored in the memory to execute the conversational knowledge management agent assistance method based on a large model according to an embodiment of the present application.

[0042] An embodiment of the present application also provides a computer-readable storage medium on which a computer program executable by a processor is stored, wherein when the program is executed by the processor, the large model-based conversational knowledge management agent assistance method provided in the embodiment of the present application is implemented.

[0043] According to the large-model-based conversation knowledge management agent assistance method and device and electronic device of the embodiment of the present application, each conversation message contained in the group message is segmented to obtain the word vector of each word contained therein, and the first correlation between the words is calculated to determine at least one topic word in the group message. At the same time, the second correlation between each conversation message and at least one conversation message adjacent in time sequence can be calculated to determine the associated messages of each conversation message, thereby forming a topic conversation group for a specific topic. For each subject conversation group, the second correlation of each word vector of the multiple conversation messages contained therein is calculated to determine the first related word of each topic word, and the third correlation between the first related word and the words adjacent in position is further calculated. In order to determine the second related word associated with it, a topic sequence for the corresponding topic word can be generated based on the first related word and the second related word. Therefore, the topic word related to multiple words can be determined from the group message containing a large number of conversation messages, so that each conversation message is re-clustered to form each topic conversation group, and then the first related word closely related to the topic word and the second related word associated with the first related word can be determined in the topic conversation group by calculating the second correlation and the third correlation, so that the expressions closely related to the topic word in the conversation message can be effectively extracted to form a topic sequence, which greatly improves the efficiency and accuracy of group message management and enables group members to obtain key information and know the to-do items more quickly.

[0044] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present application. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0046] Figure 1 A flowchart of an embodiment of the large model-based conversational knowledge management agent-assisted method provided by this application;

[0047] Figure 2 A schematic structural diagram of an embodiment of a conversational knowledge management agent assistance device based on a large model provided by this application;

[0048] Figure 3 This is a schematic structural diagram of an electronic device embodiment provided in this application. DETAILED DESCRIPTION

[0049] The following describes exemplary embodiments of the present application in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0050] Example 1

[0051] In today's information-rich world, social networking services (SNS) and online communication tools are widely used for personal and business communication. However, the sheer volume and complexity of messages in social and work groups necessitates managers to spend considerable time sifting through key information, such as to-do items, important notifications, and decision-making rationales. This information overload makes manual sifting inefficient and prone to missing important content.

[0052] Existing message management technologies have limitations when handling complex scenarios. Traditional keyword search methods fail to understand contextual semantics, resulting in inaccurate extraction results. Furthermore, group messages are often unstructured, containing a large amount of fragmented content, redundant information, and noise, further complicating information extraction.

[0053] Therefore, there is an urgent need for a large-scale model-based intelligent agent-assisted conversational knowledge management solution. Leveraging natural language processing (NLP) technology and deep learning models, this solution can extract meaningful information from massive amounts of unstructured messages. It can quickly and accurately filter out key content and structure information summarization and conclusions, thereby improving management efficiency, reducing information omissions, and meeting the needs of modern information management.

[0054] like Figure 1 As shown in Figure 1 The flowchart of an embodiment of a large model-based conversational knowledge management agent-assisted method according to an embodiment of the present application is shown. The large model-based conversational knowledge management agent-assisted method according to an embodiment of the present application may include:

[0055] S101: Obtain a first group message.

[0056] In an embodiment of the present application, a first group message may be obtained in step S101. The first group message may include multiple conversation messages sent by multiple group members. For example, an API provided by various instant messaging platforms may be used to obtain various messages in the corresponding instant messaging tools, and a group message of a specified group may be further obtained as the first group message.

[0057] The conversation message contained in the first group message may include sender information, receiver information, sending time information, sending content, etc., and in an embodiment of the present application, the sending content of the conversation message may have various data formats, for example, text data, image data, or voice data. Therefore, in an embodiment of the present application, in step S101, the data type of each conversation message in the acquired first group message may be further determined. For example, a pre-trained machine learning model may be used to calculate the correlation between each conversation message, for example, the sending content contained in the conversation message, and a preset data type, and thereby determine the data type of the sending content contained in the painting message based on the correlation, and then convert it into text data using the corresponding machine learning model according to the data type determined in this way.

[0058] For example, when it is determined that a conversation message (the content sent therein) belongs to an image data type, an image recognition model can be used in step S101 to extract at least one image feature of the conversation message, and then at least one object contained in the conversation message can be determined based on these image features. For example, it can be determined based on the image features that the conversation message or the content sent therein contains graphics of people and graphics of cars. That is, in an embodiment of the present application, a user can send pictures of people and cars taken by himself in a group as content sent, so that in step S101, when it is determined that the conversation message is of an image data type, an image recognition model can be used to extract various image features therefrom, and then determine that it contains people and cars. In an embodiment of the present application, when it is determined that the image contains a person, a face recognition model can be further used to determine the identity of the person, or when it is determined that the image contains a vehicle, a vehicle recognition model can be further used to determine the model of the vehicle, and so on. The present application is not limited to this.

[0059] When the object included in the content sent as the image data is determined, a text description corresponding to the conversation message may be further determined based on the object information of the object, and the text description may be used as text data of the conversation message.

[0060] In addition, when it is determined that the conversation message (the sent content contained therein) belongs to the voice data type, the conversation message can be converted into text using a voice recognition model in step S101, and then the text is used as the text data of the conversation message.

[0061] S102: Perform word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message.

[0062] In step S102, word segmentation processing can be performed on each conversation message in the first group message obtained in step S101. For example, the conversation message of the text data obtained in step S101 can be used as the first group message. In step S102, word segmentation processing can be directly performed on such text data to generate word vectors for the words contained in the conversation message.

[0063] In an embodiment of the present application, word embedding (WE) can be used to convert each word contained in the conversation message into a word vector, so that it can be further processed in subsequent steps. In particular, in an embodiment of the present application, sentiment-specific word embedding (SSWE) can be further used in step S102 to convert the text data obtained in step S101 into a word vector, thereby obtaining a word vector with sentiment information, which can better reflect the sentiment of the words in the text data.

[0064] S103 : Calculate first correlations between the words based on the word vectors, the corresponding position information of each word in the corresponding conversation message, and the time sequence of the corresponding conversation messages in the first group message.

[0065] S104: Determine at least one keyword in the first group message according to the first correlation.

[0066] In step S103, the first correlation between each word can be calculated for the word vectors of the words obtained after the word segmentation processing for each conversation message in step S102. Specifically, the first correlation between each word can be calculated using the word vector of each word, the position information of the word in the conversation message, and the time sequence of the conversation message in the first group message. For example, the first correlation between words can be calculated using the position information and the time sequence as the weight values of the corresponding words. The subject words in the first group message can then be determined based on the first correlation. In an embodiment of the present application, the correlations of the various words obtained in step S103 can be sorted according to the size of the values, and one or more words with the highest correlation ranking can be determined as subject words. That is, these words are associated with words in multiple conversation messages. Therefore, it is likely that when the group message was obtained in step S101, the members of the group may be discussing the topic.

[0067] S105: Calculate a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence.

[0068] S106: Determine, based on the second correlation, associated messages of each conversation message containing the topic word to form at least one topic conversation group.

[0069] After step S104, the second correlation between each conversation message contained in the first group message and the temporally adjacent conversation messages can be calculated in step S105. In step S105, the correlation between the two conversation messages can be calculated as the second correlation based on the conversation message (i.e., the sent content contained therein) as a whole. Then in step S106, the associated messages of the conversation message containing the subject word determined in step S104 can be determined based on the second correlation determined in step S105, i.e., other conversation messages that are associated with the conversation message containing the subject word. In an embodiment of the present application, when group members discuss a certain topic, they may also discuss other things in the process of the discussion. Therefore, in step S106, the correlation calculated in step S105 can be used to select messages related to the conversation message containing the subject word, i.e., such related conversation messages are likely to indicate that the group users are discussing this topic, rather than having started discussing other topics. Each topic conversation group determined in step S106 can correspond to one subject word, and can also correspond to multiple subject words. Therefore, the topic conversation group determined in step S106 may include multiple conversation messages that are continuous or discontinuous in chronological order. In other words, the conversation messages included in the group message obtained in step S101 may very likely include two or more group members discussing other matters, or even chatting. Therefore, by calculating the relevance between the conversation messages in step S105, these conversation messages unrelated to the current topic can be removed, thereby forming a topic conversation group containing only conversation messages related to the topic word or containing the topic word.

[0070] S107: For each topic conversation group in the at least one topic conversation group, calculate a second correlation of each word vector of a plurality of conversation messages contained therein.

[0071] S108: Determine first related words corresponding to each subject word based on the second relevance.

[0072] In step S107, the second correlation between the word vectors of the various words in the multiple conversation messages contained in the topic conversation group determined in step S106 can be calculated. In the embodiment of the present application, since the conversation messages in the group messages that are likely to be irrelevant to the topic have been removed through the processing of step S106, the conversation messages with a high degree of relevance to the topic that have been aggregated in this way can be recalculated in step S107, and the correlation between the words contained therein can be further calculated. In the embodiment of the present application, the second correlation can represent the semantic correlation between the word vector in a conversation message and the word vectors in other conversation messages in the topic conversation group. In step S108, the first correlation corresponding to the topic word can be determined based on such correlation. For example, the words whose calculated relevance to the topic word is higher than a preset threshold can be used as the first related words.

[0073] S109: Calculate a third correlation between the first related word and the adjacent words in the respective conversation messages.

[0074] S110: Determine a second related word associated with the first related word according to the third correlation.

[0075] S111 , generating a topic sequence for the corresponding topic word based on the first related word and the second related word.

[0076] In step S109, other related words related to the related words determined in step S108 can be further calculated. For example, when the topic discussed in the group is project A, one or more related words related to the topic word "project" can be calculated in step S108, such as progress, problem, etc. as first related words. Then, in step S109, the relevance between other words adjacent to these first related words in the positional order and the first related words can be further determined based on the first related words determined in step S108. Because during the conversation, participants are likely to continue to express the specific situation of the project progress after talking about the project progress, such specific descriptive words can be determined by calculating the relevance between the words adjacent to the first related words in position and the first related words. Therefore, in step S111, a topic sequence for the corresponding topic can be generated based on the first related words and second related words determined in this way to be displayed to relevant users, which helps users have a clearer understanding of the conversation content related to the specified topic in the group message.

[0077] In addition, in an embodiment of the present application, the conversation knowledge management agent-assisted method based on a large model can further calculate the emotions or attitudes represented by each conversation message in the subject conversation group. For example, the word vector of each word in each conversation message in the subject conversation group can be calculated to include a first result containing the position feature of the subject word corresponding to the subject conversation group in the conversation message and a second result containing the context feature of the subject word in the subject conversation group; the first result and the second result are subjected to feature fusion calculation to generate a fusion feature; the opinion tendency of the conversation message for the subject word is calculated based on the fusion feature; and the topic sequence generated in step S111 corresponding to the subject conversation group is updated based on the opinion tendency. For example, some discussion content on the topic can be screened out in the above manner, and different annotations or display effects can be given to the relevant words in the topic sequence based on the opinion tendency calculated in this way, so that users can more clearly understand the attitudes of these conversation discussions related to the topic.

[0078] Therefore, according to the conversation knowledge management agent assistance method based on a large model of an embodiment of the present application, the word vector of each word contained in the group message is obtained by segmenting each conversation message contained in the group message, and the first correlation between each word is calculated to determine at least one subject word in the group message. At the same time, the second correlation between each conversation message and at least one conversation message adjacent in time sequence can be calculated to determine the associated messages of each conversation message, thereby forming a subject conversation group for a specific subject. For each subject conversation group, the second correlation of each word vector of the multiple conversation messages contained therein is calculated to determine the first related word of each subject word, and the third correlation between the first related word and the words adjacent in position is further calculated to determine The second related words associated with it can eventually generate a topic sequence for the corresponding subject word based on the first related words and the second related words. Therefore, the subject words related to multiple words can be determined from the group message containing a large number of conversation messages, so that each conversation message can be re-clustered to form each subject conversation group, and then the first related word closely related to the subject word and the second related word associated with the first related word can be determined in the subject conversation group by calculating the second correlation and the third correlation, so that the expressions closely related to the subject word in the conversation message can be effectively extracted to form a topic sequence, which greatly improves the efficiency and accuracy of group message management and enables group members to obtain key information and know the to-do items more quickly.

[0079] Example 2

[0080] Figure 2 This is a structural diagram of an embodiment of the conversation knowledge management agent auxiliary device based on a large model provided by this application. Figure 2As shown in , the large model-based conversation knowledge management intelligent agent auxiliary device provided in the embodiment of the present application may include: an acquisition module 201, a preprocessing module 202, a first calculation module 203, a first determination module 204, a second calculation module 205, a second determination module 206, a third calculation module 207, a third determination module 208, a fourth calculation module 209, a fourth determination module 210 and a generation module 211.

[0081] The acquisition module 201 may be configured to acquire a first group message.

[0082] In an embodiment of the present application, the acquisition module 201 may acquire a first group message, which may include multiple conversation messages sent by multiple group members. For example, an API provided by various instant messaging platforms may be used to acquire various messages in the corresponding instant messaging tools, and a group message of a specified group may be further acquired as the first group message.

[0083] The conversation message contained in the first group message may include sender information, receiver information, sending time information, sending content, etc., and in an embodiment of the present application, the sending content of the conversation message may have various data formats, for example, text data, image data, or voice data. Therefore, in an embodiment of the present application, the acquisition module 201 may further determine the data type of each conversation message in the acquired first group message. For example, a pre-trained machine learning model can be used to calculate the correlation between each conversation message, for example, the sending content contained in the conversation message, and a preset data type, and thereby determine the data type of the sending content contained in the painting message based on the correlation, and then convert it into text data using the corresponding machine learning model according to the data type determined in this way.

[0084] For example, when it is determined that the conversation message (the content sent therein) belongs to the image data type, the acquisition module 201 can use the image recognition model to extract at least one image feature of the conversation message, and then determine at least one object contained in the conversation message based on these image features. For example, it can be determined based on the image features that the conversation message or the content sent therein contains graphics of people and graphics of cars. That is, in an embodiment of the present application, a user can send pictures of people and cars taken by himself in a group as content sent, so that when the acquisition module 201 determines that the conversation message is of the image data type, it can use the image recognition model to extract various image features therefrom, and further determine that it contains people and cars. In an embodiment of the present application, when it is determined that the image contains people, a face recognition model can be further used to determine the identity of the people, or when it is determined that the image contains a vehicle, a vehicle recognition model can be further used to determine the model of the vehicle, and so on. The present application is not limited to this.

[0085] When the object included in the content sent as the image data is determined, a text description corresponding to the conversation message may be further determined based on the object information of the object, and the text description may be used as text data of the conversation message.

[0086] In addition, when it is determined that the conversation message (the sent content contained therein) belongs to the voice data type, the acquisition module 201 can use the voice recognition model to convert the conversation message into text, and then use the text as the text data of the conversation message.

[0087] The pre-processing module 202 is configured to perform word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message.

[0088] The pre-processing module 202 may perform word segmentation processing on each conversation message in the first group message obtained by the obtaining module 201. For example, the conversation message of text data obtained by the obtaining module 201 may be used as the first group message. The pre-processing module 202 may directly perform word segmentation processing on such text data to generate word vectors for the words contained in the conversation message.

[0089] In this embodiment of the present application, word embedding (WE) can be used to convert each word contained in the conversation message into a word vector, so that it can be further processed in subsequent steps. In particular, in this embodiment of the present application, the pre-processing module 202 can further use sentiment-specific word embedding (SSWE) to convert the text data obtained by the acquisition module 201 into a word vector, thereby obtaining a word vector with sentiment information, which can better reflect the sentiment of the words in the text data.

[0090] The first calculation module 203 is configured to calculate a first correlation between the words based on the word vectors, the corresponding position information of each word in the corresponding conversation message, and the time sequence of the corresponding conversation messages in the first group message.

[0091] The first determining module 204 determines at least one keyword in the first group message according to the first correlation.

[0092] The first calculation module 203 can calculate the first correlation between each word based on the word vector obtained after the pre-processing module 202 performs word segmentation processing on each conversation message. Specifically, the first correlation between each word can be calculated using the word vector of each word, the position information of the word in the conversation message, and the time sequence of the conversation message in the first group message. For example, the first correlation between words can be calculated using the position information and the time sequence as the weight value of the corresponding word. The subject words in the first group message can then be determined based on the first correlation. In an embodiment of the present application, the correlations of each word obtained by the first calculation module 203 can be sorted according to the size of the value, and one or more words with the highest correlation ranking can be determined as the subject words. That is, these words are associated with words in multiple conversation messages. Therefore, it is likely that when the acquisition module 201 acquires the group message, the members of the group may be discussing the topic.

[0093] The second calculation module 205 is configured to calculate a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence.

[0094] The second determining module 206 determines, based on the second correlation, related messages of each conversation message containing the topic word to form at least one topic conversation group.

[0095] The second calculation module 205 can be used to calculate a second correlation between each conversation message included in the first group message and temporally adjacent conversation messages. The second calculation module 205 can calculate the correlation between two conversation messages based on the conversation message (i.e., the content it contains) as a whole, as the second correlation. The second determination module 206 can then, based on the second correlation determined by the second calculation module 205, determine related messages to the conversation message containing the keyword determined by the first determination module 204, i.e., other conversation messages related to the conversation message containing the keyword. In this embodiment of the present application, when group members discuss a particular topic, they may also discuss other matters during the discussion. Therefore, the second determination module 206 can use the correlation calculated by the second calculation module 205 to select messages related to the conversation message containing the keyword. In other words, such related conversation messages are likely to indicate that the group users are discussing the topic, rather than having already started discussing other topics. Each topic conversation group determined by the second determination module 206 can correspond to a single keyword, or can correspond to multiple keywords. Therefore, a topic conversation group determined by the second determination module 206 can include multiple conversation messages that are either consecutive or discontinuous in time. That is to say, the conversation messages contained in the group message obtained by the acquisition module 201 may very likely include two or more group members starting to discuss other things or even chatting. Therefore, the second calculation module 205 can be used to calculate the correlation between the conversation messages to remove these conversation messages that are not related to the current topic, thereby forming a topic conversation group that only contains conversation messages related to the topic word or containing the topic word.

[0096] The third calculation module 207 is configured to calculate, for each topic conversation group in at least one topic conversation group, a second correlation of each word vector of a plurality of conversation messages contained therein.

[0097] The third determining module 208 is configured to determine a first related word corresponding to each subject word according to the second relevance.

[0098] The third calculation module 207 can calculate the second correlation between the word vectors of each word in the multiple conversation messages contained in the topic conversation group determined by the second determination module 206. In the embodiment of the present application, since the conversation messages in the group messages that are likely to be irrelevant to the topic have been removed through the processing of the second determination module 206, the third calculation module 207 can recalculate the conversation messages with a high degree of relevance to the topic in this way, and then calculate the correlation between the words contained therein. In the embodiment of the present application, the second correlation can represent the semantic correlation between the word vector in a conversation message and the word vectors in other conversation messages in the topic conversation group. The third determination module 208 can determine the first correlation corresponding to the topic word based on such correlation. For example, the words whose calculated relevance to the topic word is higher than a preset threshold can be used as the first related words.

[0099] The fourth calculation module 209 is configured to calculate a third correlation between the first related word and the adjacent words in the respective conversation messages.

[0100] The fourth determining module 210 is configured to determine a second related word associated with the first related word according to the third correlation.

[0101] The generating module 211 is configured to generate a topic sequence for a corresponding topic word based on the first related word and the second related word.

[0102] The fourth calculation module 209 can further calculate other related words related to the related words related to the subject word determined in the third determination module 208. For example, when the topic discussed in the group is Project A, the third determination module 208 can calculate one or more related words related to the subject word "project", such as "progress", "problem", etc. as first related words. Then, the fourth calculation module 209 can further determine the relevance between other words adjacent to these first related words in the positional order and the first related words based on the first related words determined by the third determination module 208. Because during the conversation, participants are likely to continue to express the specific details of the project progress after discussing the project progress, such specific descriptive words can be determined by calculating the relevance between words adjacent to the first related words in the positional order and the first related words. Therefore, the generation module 211 can generate a topic sequence for the corresponding topic based on the first related words and second related words determined in this way, and display it to relevant users, helping users to have a clearer understanding of the conversation content related to the specified topic in the group message.

[0103] In addition, in an embodiment of the present application, the conversation knowledge management agent auxiliary device based on the large model can further calculate the emotions or attitudes represented by each conversation message in the subject conversation group. For example, the word vector of each word in each conversation message in the subject conversation group can be calculated to include a first result containing the position feature of the subject word corresponding to the subject conversation group in the conversation message and a second result containing the context feature of the subject word in the subject conversation group; the first result and the second result are subjected to feature fusion calculation to generate a fusion feature; the opinion tendency of the conversation message for the subject word is calculated based on the fusion feature; and the subject sequence generated in step S111 corresponding to the subject conversation group is updated based on the opinion tendency. For example, some discussion content on the subject can be screened out in the above manner, and different annotations or display effects can be given to the relevant words in the subject sequence based on the opinion tendency calculated in this way, so that users can more clearly understand the attitudes of these conversation discussions related to the subject.

[0104] Therefore, according to the large model-based conversation knowledge management intelligent agent auxiliary device of the embodiment of the present application, the word vector of each word contained in the group message is obtained by segmenting each conversation message contained in the group message, and the first correlation between each word is calculated to determine at least one subject word in the group message. At the same time, the second correlation between each conversation message and at least one conversation message adjacent in time sequence can be calculated to determine the associated messages of each conversation message, thereby forming a subject conversation group for a specific subject. For each subject conversation group, the second correlation of each word vector of the multiple conversation messages contained therein is calculated to determine the first related word of each subject word, and the third correlation between the first related word and the words adjacent in position is further calculated to determine The second related words associated with it can eventually generate a topic sequence for the corresponding subject word based on the first related words and the second related words. Therefore, the subject words related to multiple words can be determined from the group message containing a large number of conversation messages, so that each conversation message can be re-clustered to form each subject conversation group, and then the first related word closely related to the subject word and the second related word associated with the first related word can be determined in the subject conversation group by calculating the second correlation and the third correlation, so that the expressions closely related to the subject word in the conversation message can be effectively extracted to form a topic sequence, which greatly improves the efficiency and accuracy of group message management and enables group members to obtain key information and know the to-do items more quickly.

[0105] Example 3

[0106] The above describes the internal functions and structure of the conversational knowledge management agent assistance device based on a large model, which can be implemented as an electronic device. Figure 3 This is a schematic diagram of the structure of an electronic device embodiment provided by this application. Figure 3As shown, the electronic device includes a memory 31 and a processor 32 .

[0107] Memory 31 is used to store programs. In addition to the aforementioned programs, memory 31 may also be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, contact data, phone book data, messages, images, videos, etc.

[0108] The memory 31 can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0109] Processor 32 is not limited to a central processing unit (CPU) but may also be a graphics processing unit (GPU), a field programmable gate array (FPGA), an embedded neural network processor (NPU), or an artificial intelligence (AI) chip. Processor 32 is coupled to memory 31 and executes a program stored in memory 31. When executed, the program implements the large model-based conversational knowledge management agent-assisted method of the first embodiment.

[0110] Further, if Figure 3 As shown, the electronic device may further include: a communication component 33, a power component 34, an audio component 35, a display 36 and other components. Figure 3 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 3 Components shown.

[0111] The communication component 33 is configured to facilitate wired or wireless communication between the electronic device and other devices. The electronic device can access a wireless network based on a communication standard, such as WiFi, 3G, 4G or 5G, or a combination thereof. In an exemplary embodiment, the communication component 33 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 33 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0112] The power supply assembly 34 provides power to various components of the electronic device. The power supply assembly 34 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device.

[0113] The audio component 35 is configured to output and / or input audio signals. For example, the audio component 35 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 31 or transmitted via the communication component 33. In some embodiments, the audio component 35 also includes a speaker for outputting audio signals.

[0114] The display 36 includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can detect not only the boundaries of a touch or slide action, but also the duration and pressure associated with the touch or slide operation.

[0115] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A conversational knowledge management agent-assisted method based on a large model, characterized in that: include: Obtaining a first group message, wherein the first group message includes a plurality of conversation messages sent by a plurality of group members; Performing word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message; Calculating a first correlation between the words based on the word vectors, position information of each corresponding word in the corresponding conversation message, and a time sequence of the corresponding conversation messages in the first group message; determining, based on the first correlation, at least one topic word in the first group message; Calculating a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence; Determining, based on the second correlation, associated messages of each conversation message containing the keyword to form at least one topic conversation group, wherein each topic conversation group corresponds to a topic word and includes a plurality of conversation messages that are continuous or discontinuous in time sequence; For each topic conversation group in the at least one topic conversation group, calculating a second correlation between word vectors of the plurality of conversation messages contained therein, wherein the second correlation represents a semantic correlation between a word vector in a conversation message and word vectors in other conversation messages in the topic conversation group; Determining a first related word corresponding to each subject word based on the second relatedness; Calculating a third correlation between the first related word and words adjacent in position in the respective conversation messages; determining, based on the third correlation, a second related word associated with the first related word; A topic sequence for corresponding topic words is generated based on the first related words and the second related words.

2. The method for assisting conversational knowledge management agents based on a large model according to claim 1, characterized in that: The session message includes sender information, receiver information, sending time information, and sending content.

3. The method for assisting conversational knowledge management agents based on a large model according to claim 1, characterized in that: The acquiring of the first group message includes: Determining the data type of each conversation message in the acquired first group message; The data type is converted into text data using a corresponding machine learning model.

4. The method for assisting conversational knowledge management agents based on a large model according to claim 3, characterized in that: Converting the data into text data using a corresponding machine learning model according to the data type includes: When it is determined that the conversation message belongs to the image data type, extracting at least one image feature of the conversation message using an image recognition model; determining at least one object included in the conversation message according to the image feature; Determining a text description corresponding to the conversation message according to the object information of the object; The text description is used as the text data.

5. The method for assisting conversational knowledge management agents based on a large model according to claim 3, characterized in that: Converting the data into text data using a corresponding machine learning model according to the data type includes: When it is determined that the conversation message belongs to the voice data type, converting the conversation message into text using a speech recognition model; The character text is used as the text data.

6. The method for assisting conversational knowledge management agents based on a large model according to claim 1, characterized in that: The method further comprises: For each word in each conversation message in the topic conversation group, a word vector is calculated, which includes a first result containing a position feature of the topic word corresponding to the topic conversation group in the conversation message and a second result containing a context feature of the topic word in the topic conversation group; Performing feature fusion calculation on the first result and the second result to generate a fusion feature; Calculating the opinion tendency of the conversation message regarding the keyword based on the fusion feature; and The topic sequence corresponding to the topic conversation group is updated according to the opinion tendency.

7. A conversational knowledge management agent auxiliary device based on a large model, characterized in that: include: An acquisition module, configured to acquire a first group message, wherein the first group message includes a plurality of conversation messages sent by a plurality of group members; a preprocessing module, configured to perform word segmentation processing on each conversation message included in the first group message to generate a word vector for each word included in each conversation message; a first calculation module, configured to calculate a first correlation between the words based on the word vectors, position information of each word in the corresponding conversation message, and a time sequence of the corresponding conversation messages in the first group message; a first determining module, configured to determine at least one keyword in the first group message based on the first correlation; a second calculation module, configured to calculate a second correlation between each conversation message included in the first group message and at least one conversation message adjacent to it in time sequence; a second determining module configured to determine, based on the second correlation, associated messages of each conversation message containing the keyword to form at least one topic conversation group, wherein each topic conversation group corresponds to a topic word and includes a plurality of conversation messages that are continuous or discontinuous in chronological order; a third calculation module, configured to calculate, for each topic conversation group in the at least one topic conversation group, a second correlation between word vectors of the plurality of conversation messages contained therein, wherein the second correlation represents a semantic correlation between a word vector in a conversation message and word vectors in other conversation messages in the topic conversation group; a third determining module, configured to determine a first related word corresponding to each subject word according to the second relevance; a fourth calculation module, configured to calculate a third correlation between the first related words and words adjacent to them in respective conversation messages; a fourth determining module, configured to determine a second related word associated with the first related word according to the third correlation; A generating module is used to generate a topic sequence for the corresponding topic word based on the first related word and the second related word.

8. The large model-based conversational knowledge management agent auxiliary device according to claim 7, characterized in that: The acquisition module is further configured to: Determining the data type of each conversation message in the acquired first group message; The data type is converted into text data using a corresponding machine learning model.

9. An electronic device, characterized in that: include: Memory, used to store programs; A processor, configured to run the program stored in the memory to execute the large model-based conversational knowledge management agent assistance method as described in any one of claims 1 to 6.

Images