Group chat message processing method and device, equipment and storage medium

Abstract

The invention provides a group chat message processing method and device, equipment and a storage medium, relates to the technical field of computers, in particular to the fields of information flow, artificial intelligence, large models and the like, and can be applied to application scenes of intelligent assistants, virtual assistants, intelligent e-commerce and the like. According to the specific implementation scheme, a to-be-processed message in a chat group is acquired; based on the context information of the to-be-processed message, determining an emotion category of the to-be-processed message and a model field required by a preset message model; matching the emotion category and the model field of the to-be-processed message with a preset message model set of the target object; under the condition that any preset message model in the preset message model set is matched, determining a reminding mode of the to-be-processed message; and performing message reminding on the to-be-processed message based on the reminding mode. In the embodiment of the invention, the important information in the group chat is identified, captured and reminded, so that various troubles brought to the user due to information overload are reduced.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to the fields of information flow, artificial intelligence, and large models, and can be used in application scenarios such as intelligent assistants, virtual assistants, and intelligent e-commerce. Background Art

[0002] With the development of computer technology, instant messaging software has become an indispensable part of people's lives, among which group chat messages have become one of the widely used communication tools.

[0003] Group chat can effectively promote communication and information sharing among team or community members, improve the efficiency of decision-making and problem solving. It also supports easy initiation of discussions, sharing of diverse content, and helps to enhance cohesion among people. Summary of the invention

[0004] The present disclosure provides a method, apparatus, device and storage medium for processing a group chat message.

[0005] According to one aspect of the present disclosure, a method for processing a group chat message is provided, comprising:

[0006] Get pending messages in the chat group;

[0007] Based on the context information of the message to be processed, determine the sentiment category of the message to be processed and the model fields required by the preset message model;

[0008] Match the sentiment category and model field of the message to be processed with the preset message model set of the target object;

[0009] When any preset message model in the preset message model set is matched, determining a reminder method for the message to be processed;

[0010] Prompts the processing message based on the reminder method.

[0011] According to another aspect of the present disclosure, a device for processing a group chat message is provided, comprising:

[0012] The acquisition module is used to obtain the pending messages in the chat group;

[0013] A first determination module is used to determine the emotion category of the message to be processed and the model fields required by the preset message model based on the context information of the message to be processed;

[0014] A matching module, used to match the sentiment category and model field of the message to be processed with the preset message model set of the target object;

[0015] A second determination module is used to determine a reminder method for the message to be processed when any preset message model in the preset message model set is matched;

[0016] The reminder module is used to remind the messages to be processed based on the reminder method.

[0017] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0018] at least one processor; and

[0019] a memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.

[0021] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.

[0022] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements any method according to the embodiments of the present disclosure when executed by a processor.

[0023] In the disclosed embodiment, by intelligently analyzing and screening the group chat content, users are assisted in identifying, capturing and reminding important information in the group chat, thereby reducing the various troubles caused to users by information overload.

[0024] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0026] Figure 1 is a flowchart of a method for processing a group chat message according to an embodiment of the present disclosure;

[0027] Figure 2 It is a flowchart of determining the emotion category of a message to be processed and the model fields required by a preset message model according to an embodiment of the present disclosure;

[0028] Figure 3 is a schematic diagram of a structural flow of processing a message to be processed according to an embodiment of the present disclosure;

[0029] Figure 4 It is a flowchart of building a preset message model based on group messages according to an embodiment of the present disclosure;

[0030] Figure 5 is a schematic diagram of a process of constructing a preset message model based on keywords according to an embodiment of the present disclosure;

[0031] Figure 6 It is a schematic diagram of the overall process of constructing a preset message model based on configuration information according to an embodiment of the present disclosure;

[0032] Figure 7 It is a schematic diagram of the overall structural flow of a method for processing a group chat message according to an embodiment of the present disclosure;

[0033] Figure 8 is a structural diagram of a device for processing group chat messages according to an embodiment of the present disclosure;

[0034] Fig. 9 It is a block diagram of an electronic device used to implement the method for processing group chat messages according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0036] With the continuous development of information technology, instant messaging software has become an indispensable communication tool in daily life and work. In particular, the group chat function in instant messaging software is widely used in multi-person discussion and cooperation scenarios. However, with the expansion of group chat scale and the increase in the number of messages, the problems of information redundancy and overload have become increasingly prominent. Users often find it difficult to quickly filter out valuable content from massive messages, which not only increases the difficulty of information processing, but also may cause key information to be ignored or even missed.

[0037] In view of this, the embodiment of the present disclosure provides a method for processing group chat messages. The method assists users in identifying, capturing and reminding important information in group chats by intelligently analyzing and screening the content of group chats, thereby reducing the various troubles caused to users by information overload.

[0038] like Figure 1 FIG. 1 is a flow chart of a method for processing a group chat message proposed in an embodiment of the present disclosure, including:

[0039] S101, obtaining the pending messages in the chat group.

[0040] All messages in the chat group can be obtained in real time as pending messages through the API (Application Programming Interface) interface of the instant messaging software. These messages may include text, pictures, links, voice and other forms.

[0041] It should be noted that all information involving user privacy is collected in strict compliance with relevant laws and policies and with the user's knowledge and consent. While protecting the security of the user's personal data, it will not be used for any unauthorized purpose.

[0042] S102, based on the context information of the message to be processed, determining the emotion category of the message to be processed and the model fields required by the preset message model.

[0043] Based on the context message of the pending message, using Natural Language Processing (NLP) technology, the emotional category expressed by the pending message can be analyzed, such as positive, neutral, negative, etc., according to the content, tone, and wording of the pending message and its context message. At the same time, the key information in the pending message can be extracted as the model field required by the preset message model. This key information is used to remind the target object (i.e., the user) and can be customized by the target object or automatically mined. This key information may include specific keywords, such as sender, recipient, task name, time period, subject, etc.

[0044] S103, matching the emotion category and model field of the message to be processed with the preset message model set of the target object.

[0045] The preset message model is a model used to describe the target object's basic preference for group chat reminders. These preset message models in the preset message model set are mined based on historical data and specific algorithms. The preset message model set may contain multiple message models, each corresponding to its own processing logic and reminder method. During implementation, these preset models can be matched according to the sentiment category and model field of the message to find the most appropriate processing strategy.

[0046] S104: When any preset message model in the preset message model set is matched, a reminder method for the message to be processed is determined.

[0047] Once a matching preset message model is found, the message reminder method can be determined according to the rules of the model. For example, for messages that are judged to be urgent and important, the target object may be immediately reminded through a pop-up window and a ringtone; while for ordinary messages, only a small reminder may be given in the message bar.

[0048] S105: Prompt the message to be processed based on the reminder method.

[0049] That is, according to the determined reminder method, different degrees and types of message reminder methods are provided to the target object, and the message is pushed. While ensuring the best reminder effect, the interference is minimized to enhance the experience of the target object.

[0050] In the disclosed embodiment, by deeply analyzing the context information of the message to be processed and determining the sentiment category and model field of the message, the message content can be considered more comprehensively and the possibility of misjudgment can be reduced. By matching the preset message model, a personalized reminder method can be provided for different message contents to reduce or avoid unnecessary reminders, thereby preventing the target object from being disturbed by irrelevant information, so that the target object can more efficiently obtain and process information related to itself.

[0051] In the embodiment of the present disclosure, based on the context information of the message to be processed, the emotion category of the message to be processed and the model fields required by the preset message model are determined, such as Figure 2 As shown, it may include the following:

[0052] S201, using a feature extractor to extract target features from the message to be processed and context information.

[0053] A feature extractor is a mechanism in natural language processing that can extract features from raw data that are helpful for classification, clustering or other tasks, so that the data can be better understood and processed in subsequent steps.

[0054] In the disclosed embodiment, the feature extractor includes an encoder, a multi-head self-attention mechanism network module and a hidden layer.

[0055] Among them, the encoder is responsible for encoding the input message to be processed and the context information to convert the original information into a form that is easier for the model to understand and process, providing a basis for subsequent feature extraction.

[0056] The multi-head self-attention mechanism network module is a network based on the self-attention mechanism, which can simultaneously consider the information of all positions in the input sequence, so as to accurately capture the dependencies and important features in the sequence in combination with contextual information, thereby further enhancing the expressiveness of the model and the accuracy of feature extraction.

[0057] The hidden layer is located after the multi-head self-attention network module in the feature extractor. The core semantic features output by the multi-head self-attention network module are further processed and transformed to obtain the final target features.

[0058] During implementation, before using the feature extractor to extract the target features, in order to improve the extraction efficiency of the target features, the message to be processed and the context information may be preprocessed. Preprocessing includes data cleaning to achieve denoising. Preprocessing may also include steps such as word segmentation, stop word removal, stem extraction, etc. of the information after data cleaning to convert the original text into a format that can be processed by the feature extractor. The feature extractor in the disclosed embodiment may select pre-trained language models such as BERT (Bidirectional Encoder Representations from Transformers, based on the bidirectional representation of transformer encoders) and RoBERTa (Bidirectional Encoder Representations from Transformers, based on the bidirectional representation of transformer encoders). These models can encode the message to be processed and its context to generate a feature vector containing rich context information. These feature vectors can capture key information such as semantics, syntax, and contextual relationships in the text.

[0059] Based on the aforementioned network structure, the specific implementation steps of using a feature extractor to extract target features from the message to be processed and the context information are as follows:

[0060] In step A1, an encoder is used to encode the message to be processed and the context information to obtain encoding features; the encoding features are used to model the semantics and context environment of the message to be processed and the context information.

[0061] In step A2, a multi-head self-attention network module is used to process the encoded features to parse out the core semantic features in the message to be processed and the context information.

[0062] In step A3, the hidden layer is used to process the core semantic features to obtain the target features.

[0063] The hidden layer is used to further process and transform the core semantic features to obtain the final target features. The target features are the output results of the feature extractor and will be used for subsequent task processing. In the process of generating target features, the hidden layer converts the core semantic features into a form that is more suitable for subsequent task processing through a series of nonlinear transformations and mapping operations.

[0064] In the disclosed embodiment, an encoder is used to encode the message to be processed and the context information to obtain the encoded features, which can comprehensively capture the semantic information in the message and the context, as well as the relationship between them, and provide rich semantic representations for subsequent processing. The multi-head self-attention mechanism network module processes the encoded features and can parse out the core semantic features in the message to be processed and the context information, thereby improving the accuracy and efficiency of the processing. The hidden layer further processes the core semantic features to obtain the target features, which helps to extract deeper feature representations and provide strong support for subsequent tasks (such as classification, model field recognition, etc.). In summary, the process of extracting target features from the message to be processed and the context information using a feature extractor can significantly improve the system's semantic understanding ability, processing accuracy and overall performance through the collaborative work of the encoder, the multi-head self-attention mechanism network module and the hidden layer.

[0065] S202, input the target feature into the sentiment classification head to obtain the sentiment category of the message to be processed.

[0066] S203, input the target features into the key information classification header to obtain the model field.

[0067] like Figure 3 As shown, in S301, the pre-processed message to be processed and its context information are obtained. In S302, the message to be processed and its context information are segmented and converted into a message that can be used by the encoder ( Figure 3 The figure shows the word vector processed by the encoder built based on several Transformer encoding layers. For example, the word segmentation tool is used to tokenize the text, and then the token is converted into a word embedding. Then in S303, through several layers of Transformer (neural network architecture for processing sequence data) encoding layers in the encoder, each token is gradually integrated with its global information in the context to generate encoding features to model the context-related semantic representation and context.

[0068] In S304, a multi-head self-attention mechanism network module is used to process the encoded features to obtain core semantic features. Figure 3 As shown in the figure, through the multi-head attention mechanism, the corresponding attention weight is calculated for each Token, and the attention calculation is completed in multiple projection spaces to ensure fine-grained context understanding.

[0069] The hidden layer collects the context information of each token and is the input for subsequent analysis. Figure 3 As shown, a multi-head self-attention mechanism network module is used to process the encoded features to obtain core semantic features. In S305, the target features used as input for sentiment classification and key information classification can be obtained through hidden layer processing.

[0070] like Figure 3 As shown, in S306, the target features obtained from the feature extractor are input into the sentiment classification head to obtain the sentiment category.

[0071] Among them, the sentiment classification head is a trained classifier. By connecting a fully connected network layer on the semantic vector, it completes the classification of the input feature vector, outputs the sentiment category of the message to be processed, such as positive, negative, neutral, etc., and performs sentiment labeling on it.

[0072] At the same time, in S307, the target features obtained from the feature extractor are input into the key information classification head to obtain the model field.

[0073] Similar to the sentiment classification head, the key information classification head includes a fully connected layer and a discriminator. By introducing the discriminator on the fully connected layer, the corresponding category is output according to the content of the context semantic vector, such as tasks, discussions, decisions, key events, etc.

[0074] Among them, the fully connected layer is responsible for processing the target features, and converts the target features into features to be discriminated through a series of weight and bias calculations. The discriminator is used to receive the features to be discriminated, and output the weights of each word segmentation and determine the urgency of the message to be processed.

[0075] During implementation, the target features are input into the key information classification head, which analyzes the global semantic understanding vector, extracts important keywords, field-related information, tasks, and decisions as key fields, and annotates them according to their importance. The specific implementation steps are as follows:

[0076] Step B1, processing the target features based on the fully connected layer to obtain the features to be discriminated.

[0077] As mentioned above, after the target features are input into the fully connected layer, the fully connected layer can process the target features through weighted summation and nonlinear activation function to obtain the features to be discriminated. These features have been preliminarily processed and transformed and are ready to enter the discriminator for further analysis.

[0078] Step B2: input the discriminant features into the discriminator to obtain the weight of each word in the message to be processed, and determine the urgency of the message to be processed.

[0079] The features to be discriminated are input into the discriminator, which calculates the weight of each word segmentation based on its internal learning rules and parameters. At the same time, the discriminator also determines the urgency of the message to be processed based on the overall features.

[0080] Step B3: identifying key fields from the message to be processed based on the weights of each word segment.

[0081] According to the weights of each word output by the discriminator, by setting a certain threshold or sorting algorithm, it is possible to identify which words are more important in the message to be processed, and select them from these words as key fields. Key fields refer to words or phrases that are important for understanding the content of the message, performing tasks, or making decisions.

[0082] The identified key fields, together with the urgency determined by the discriminator, constitute the model field. The model field is the final output of the key information classification head. It not only includes the key information in the message to be processed, but also reflects the urgency of the message, providing an important basis for subsequent message reminders.

[0083] In the disclosed embodiment, the weights of different word segments in the message to be processed are obtained by the discriminator, so that the model can process diverse input data. Through the combination of the fully connected layer and the discriminator, the target features can be processed more accurately, thereby determining the urgency of the message to be processed and extracting the key fields in the message to be processed. This reduces the error in information processing and improves the accuracy of key information identification. The target object no longer needs to manually filter a large amount of information and can directly obtain important content, thereby improving the target object experience.

[0084] It should be noted that the execution timing of step S202 and step S203 is not restricted, and the embodiment of the present disclosure does not limit this.

[0085] In the disclosed embodiment, by using a feature extractor to extract target features from the message to be processed and its context information, the meaning and background of the message to be processed can be more fully understood. By inputting these target features into the sentiment classification head, the sentiment category of the message to be processed can be more accurately determined. By inputting the target features into the key information classification head, the model fields required for the preset message model can be accurately identified, which can ensure that the model can focus on key information when processing messages, thereby improving the overall processing efficiency and accuracy, thereby improving the target object experience and satisfaction.

[0086] In different scenarios, the important messages that the target object is concerned about will also change accordingly. For example, in the case of a work group, among the large amount of information received by the target object, the instructions of the leader, the assignment of tasks and other information are particularly important; in the case of a parent group, the important messages in the group chat are important school notices or information related to children's learning and activities; in the case of a shopping group, the target object is more concerned about discount information related to the product.

[0087] Therefore, for different application scenarios and different target groups, you can create preset message models to meet their different message reminder needs. Figure 4FIG. 1 is a flow chart of creating a preset message model in an embodiment of the present disclosure, including:

[0088] S401, obtaining historical group chat messages and the target object's interaction records with the historical group chat messages.

[0089] By connecting to the APIs of multiple instant messaging software platforms, historical messages are collected from group chat records. After obtaining historical group chat messages, it is necessary to further collect the target object's interaction records on these messages. These interaction records usually include but are not limited to:

[0090] Reading time: refers to the reading time of some messages. If a message is read repeatedly, the reading time of the message will increase, indicating that the target audience pays attention to the message.

[0091] Notification click rate: The percentage of people who actually click on the notification to view the message content when they receive an instant messaging software reminder message.

[0092] Message reply frequency: This refers to the number of replies sent by the target to the instant messaging software reminder messages received.

[0093] Message reply time: the time from receiving the reminder message from the instant messaging software to sending the reply.

[0094] Interactive feedback: feedback from the target object after the message reminder is made based on the method provided in the embodiment of the present disclosure.

[0095] S402, inputting historical group chat messages and interaction records into the big model, so that the big model analyzes and summarizes multiple description statements.

[0096] In computer science, large models usually refer to models with a large number of parameters or complex structures.

[0097] Among them, the large model can be a large language model (LLMs). Large language models refer to a specific type of large model that is specifically used to process text data. This type of model is a natural language processing model based on a neural network, which can be used to generate, understand and process text data. Large language models can have tens of billions of parameters, can generate high-quality text, and can be used for various natural language processing tasks, such as question answering, text generation, dialogue systems, etc.

[0098] The large language model has good reasoning ability and small sample learning ability. The large language model can understand the model on a large scale or can be trained on a large number of samples. Based on the large model, accurate semantic understanding can be achieved.

[0099] The collected historical group chat messages and interaction records are used as input and passed to the big model for analysis. The big model can use its long text analysis and summary capabilities to understand the contextual relationship based on the input text data and generate relevant descriptive sentences. These descriptive sentences are used to describe the messages with which characteristics the target object is more concerned. For example, it can cover the theme, emotional tendency, key information, etc. of the group chat. For example, based on the high response rate of the target object to the reminder message of contact A in the group chat, the big model can generate multiple descriptive sentences, such as "the target object and contact A have XXX work task association", "the target object pays special attention to the words "please reply" in the message of contact A", "the target object controls the message reply time of contact A within one minute", etc. Based on this type of related description, the key information can be extracted for the subsequent step S403 to find the group chat message related to this type of description sentence.

[0100] S403, obtaining group chat messages corresponding to the multiple description sentences in historical group chat messages.

[0101] Based on the description sentences generated by the large model, the corresponding specific messages are searched in the historical group chat messages through text matching and retrieval technologies, such as keyword matching and semantic similarity calculation.

[0102] S404: construct a preset message model based on the group chat messages corresponding to the multiple description statements.

[0103] Based on the found group chat messages corresponding to the description sentences, a preset message model is constructed. For example, fields such as the subject of interest and the topic can be extracted from the group chat messages corresponding to the multiple description sentences to construct the fields included in the preset message model.

[0104] In the disclosed embodiment, by analyzing historical group chat messages and the target object's interaction records with historical group chat messages through a large model, group chat messages that are likely to attract the target object's attention and interaction can be more accurately identified, so that the target object's interests and needs can be analyzed based on the group chat messages, and a more personalized and highly relevant preset message model can be created. By creating a preset message model, the target object can more accurately receive the group chat messages it needs, thereby improving the information acquisition rate and increasing the target object's experience.

[0105] In the embodiment of the present disclosure, it is also possible to automatically mine corresponding keywords from group messages by using NLP and text mining algorithms, and create a preset message model based on the keywords. The specific implementation method is as follows: Figure 5 As shown, including the following:

[0106] S501, mining keywords from the chat group based on keyword mining technology.

[0107] Collect historical chat records in the chat group as a data source, and use the NLP model to automatically extract key information. For example, the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm in natural language processing technology is used to calculate the word frequency to extract important, frequently appearing words in the message context but uncommon in the global data as potential keywords.

[0108] For complex scenarios or those that require processing large amounts of data, you can use context-based deep learning models such as Word2Vec (a learning model), or use word vector algorithms to extract high-value words in the context as keywords.

[0109] S502: Construct a preset message model based on keywords.

[0110] Based on mining keywords from chat groups, a preset message model is constructed.

[0111] In the disclosed embodiment, the keyword mining technology can quickly identify key information from a large amount of chat content, reduce the time cost of the target object to read a large amount of messages, and improve the efficiency of information processing, thereby improving the target object's satisfaction and experience in the process of using instant messaging software.

[0112] In order to meet the personalized needs of the target object, in some group chat or task scenarios, the target object is supported to customize the preset message model. For example, the target object can manually set specific keywords. These keywords are often key information that is important to the scenario, such as key figures or tags related to project cooperation and approval process.

[0113] Therefore, in the embodiment of the present disclosure, a preset message model can also be created according to the requirements of the target object, such as Figure 6 As shown, including:

[0114] S601, responding to configuration information input by a target object on a message setting page.

[0115] On the message settings page of the instant messaging software, the target object can set the corresponding configuration information according to their own needs. For example, set the type, format, sending time, and sending object of the reminder message; or manually set specific keywords to set key figures or tags related to project cooperation and approval process as specific keywords. Once the configuration information is entered, the information can be collected and parsed to understand the configuration intention of the target object.

[0116] In some embodiments, based on excellent data analysis methods, active analysis and summary can be performed based on reports in the task group, thereby realizing automated task management and tracking task processing progress.

[0117] During implementation, the target object can analyze and organize messages related to the target task through corresponding voice commands. Through NLP technology, key features such as task name, participants, stage, completion status, etc. are automatically extracted from the collected messages to complete task tracking and progress statistics.

[0118] S602: Construct a preset message model based on the configuration information.

[0119] Based on the parsed configuration information, the system will build a preset message model, which can be used to identify the target object's preference settings in the subsequent use process and provide message reminders. For example, the target object sets "**" delivery", "meeting minutes", "approval" and other keywords.

[0120] After extracting these keywords, help the target object focus on tracking these custom keywords. When these keywords are identified, send different reminder messages to the target object.

[0121] In the disclosed embodiment, the target object is allowed to input configuration information on the message setting page, so that the preset message model can be customized according to the actual needs of the target object. By constructing the preset message model based on the input configuration information, the reminder message can be more in line with the expectations of the target object.

[0122] In addition, since the preset message model is built based on custom configuration information, processing group chat information based on the preset message model can flexibly adapt to different target object needs and business scenarios. As the needs of target objects continue to change and business scenarios continue to expand, the preset message model can also be easily adjusted and optimized to meet new needs and scenarios.

[0123] In the disclosed embodiment, the personalized preference settings of the target object are taken into consideration. For example, the target object may wish to set messages from a specific sender or containing specific keywords as high priority reminders. Therefore, for the pending messages, the reminder level of the pending messages can be further determined according to the preference settings of the target object.

[0124] During implementation, messages can be divided into three levels of importance: low, medium, and high, based on the target object's preferences, such as specific keywords, senders, content types, or time periods. The higher the reminder level, the more important or urgent the message is, and needs to be processed as soon as possible. For example, negative emotions in the group, such as "when will the ** work be processed?" and "** tasks have not been handed over," can be set with a higher priority so that the technology can pay attention to such urgent matters.

[0125] Afterwards, according to the reminder level of the message to be processed, the reminder method corresponding to the reminder level can be filtered in the reminder method set.

[0126] The reminder method set is a set of pre-set notification methods, such as sound reminder, vibration reminder, pop-up reminder, email notification, SMS notification, etc. These reminder methods can be customized according to different devices and instant messaging software platforms to meet the needs of the target objects in different scenarios.

[0127] According to the reminder level of the pending message, the system will filter out the corresponding reminder method from the reminder method set. Generally, high-priority messages will use more direct and eye-catching reminder methods, such as sound reminders and pop-up reminders; while low-priority messages may use gentler reminder methods, such as email notifications or SMS notifications.

[0128] In the disclosed embodiment, by setting reminder levels for messages to be processed and selecting the most suitable reminder method, it is possible to ensure that the target object can receive important messages in a timely manner, while avoiding excessively disturbing the target object with less urgent messages, thereby improving the target object's user experience. For target objects in work scenarios, this reminder method can help the target object prioritize the most important tasks and messages, thereby effectively improving work efficiency.

[0129] In some embodiments, when selecting a reminder method, the current state of the target object may also be considered. For example, if the target object is in a meeting or in a silent environment, the use of a sound reminder may be avoided, and a vibration reminder or an email notification may be selected instead.

[0130] The target audience can personalize the reminder mode according to their needs and preferences. For example, the target audience can set to receive only certain types of reminders during certain time periods, or turn off certain reminder modes completely.

[0131] In the embodiment of the present disclosure, in order to meet the specific needs and preferences of the target object to the greatest extent, the preset message model set can be updated based on the feedback of the target object on the message to be processed, and finally the message reminder strategy is optimized.

[0132] During implementation, feedback data from the target object on the messages to be processed may be collected, including but not limited to indicators such as message viewing rate, message reply interval, message feedback, etc. These feedback data directly reflect the target object's acceptance and preference for the message content.

[0133] Using these feedback data as samples, the big model can analyze and summarize which preset message models in the preset message model set the target object is interested in and which ones have reduced interest, thereby updating the preset message model set and the reminder method corresponding to the preset message model.

[0134] In the disclosed embodiment, the preset message model set is updated based on the target object's feedback on the processed message. That is, as the target object's feedback accumulates, the new data is continuously learned and adapted to more accurately understand the target object's intentions and needs, thereby adjusting the preset message model to ensure that it can keep pace with the times and effectively meet the needs of the target object.

[0135] In summary, the method for processing group chat messages in the embodiment of the present disclosure is as follows: Figure 7 As shown:

[0136] S701, obtaining group chat messages in real time by connecting to an interface of at least one instant messaging software platform.

[0137] S702: For different application scenarios, the target object can build a preset message model by customizing keywords.

[0138] S703, when the target object does not specify keywords, keyword mining technology can be used to mine keywords, so as to mine relatively important keywords from the obtained group chat messages for building a preset message model.

[0139] S704, a corresponding description statement may also be generated through a large model according to the target object's interaction record with the group chat message, the corresponding group chat message may be found through the description statement, and a preset message model may be constructed according to the found group chat message.

[0140] S705, pre-processing the group chat message by means of content cleaning, labeling, etc.

[0141] S706, combining the context of the group chat message, classifying the sentiment and key information, and obtaining the corresponding sentiment category and key information level (i.e., model field). Based on the obtained sentiment category and key information level, matching with the preset message model set of the target object is performed to obtain a matching preset message model.

[0142] S707, when a preset message model is matched, the importance of the message is determined based on pre-built preset rules, and a corresponding reminder method is adopted to remind the target object.

[0143] The preset rule, for example, if the sentiment category is negative and the urgency of the message in the key information level is high, the most significant message reminder method is used for reminder.

[0144] S708, continuously optimizing the preset message model according to the target object's feedback information on the reminder message.

[0145] Based on the above-mentioned group chat message processing method, based on the personalized configuration of the target object, different types of message reminders can be provided by matching different preset message models to ensure that the target object can capture key information in a timely manner.

[0146] Based on the same technical concept, the embodiment of the present disclosure also provides a group chat message processing device 800, such as Figure 8 As shown, including:

[0147] The acquisition module 801 is used to acquire the pending messages in the chat group;

[0148] A first determination module 802 is used to determine the emotion category of the message to be processed and the model fields required by the preset message model based on the context information of the message to be processed;

[0149] A matching module 803 is used to match the emotion category and model field of the message to be processed with a preset message model set of the target object;

[0150] The second determination module 804 is used to determine the reminder mode of the message to be processed when any preset message model in the preset message model set is matched;

[0151] The reminder module 805 is used to remind the message to be processed based on the reminder method.

[0152] In some embodiments, the first determining module includes:

[0153] An extraction unit, used for extracting target features from the message to be processed and the context information using a feature extractor;

[0154] A first classification unit is used to input the target feature into the sentiment classification head to obtain the sentiment category of the message to be processed; and

[0155] The second classification unit is used to input the target features into the key information classification head to obtain the model field.

[0156] In some embodiments, the feature extractor includes an encoder, a multi-head self-attention mechanism network module, and a hidden layer; an extraction unit, specifically used for:

[0157] An encoder is used to encode the message to be processed and the context information to obtain encoding features; the encoding features are used to model the semantics and context environment of the message to be processed and the context information;

[0158] A multi-head self-attention network module is used to process the encoded features to parse out the core semantic features in the message to be processed and the context information;

[0159] The hidden layer is used to process the core semantic features to obtain the target features.

[0160] In some embodiments, the key information classification head includes a fully connected layer and a discriminator; the second classification unit is specifically used to:

[0161] Process the target features based on the fully connected layer to obtain the features to be discriminated;

[0162] Input the discriminant features into the discriminator to obtain the weight of each word in the message to be processed, and determine the urgency of the message to be processed;

[0163] Identify key fields from the message to be processed based on the weights of each word segmentation;

[0164] Among them, the model fields include urgency and key fields.

[0165] In some embodiments, the method further comprises a first building block for:

[0166] Obtain historical group chat messages and the target object's interaction records with historical group chat messages;

[0167] Input historical group chat messages and interaction records into the big model so that the big model can analyze and summarize multiple descriptive statements;

[0168] Obtain group chat messages corresponding to multiple description statements in historical group chat messages;

[0169] Based on the group chat messages corresponding to multiple description sentences, a preset message model is constructed.

[0170] In some embodiments, the method further comprises a second building block for:

[0171] Mining keywords from chat groups based on keyword mining technology;

[0172] Build preset message models based on keywords.

[0173] In some embodiments, a third building block is further included for:

[0174] Responding to configuration information input by the target object on the message setting page;

[0175] Based on the configuration information, a preset message model is constructed.

[0176] In some embodiments, the second determining module is used to:

[0177] Determine the alert level for pending messages;

[0178] In the reminder method set, filter the reminder method corresponding to the reminder level.

[0179] In some embodiments, a feedback module is further included for:

[0180] Based on the target object's feedback on the message to be processed, the preset message model set is updated.

[0181] For the description of specific functions and examples of each module and submodule of the device in the embodiment of the present disclosure, reference can be made to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0182] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0183] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0184] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0185] like Fig. 9 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0186] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0187] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as the processing method of the group chat message. For example, in some embodiments, the processing method of the group chat message may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the processing method of the group chat message described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the processing method of the group chat message in any other appropriate manner (e.g., by means of firmware).

[0188] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0189] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0190] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0191] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0192] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0193] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0194] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0195] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for processing a group chat message, comprising: Get pending messages in the chat group; Determine the emotion category of the message to be processed and the model fields required by the preset message model based on the context information of the message to be processed; Matching the emotion category and model field of the message to be processed with a preset message model set of the target object; In the case of matching any preset message model in the preset message model set, determining a reminder method for the message to be processed; A message reminder is performed on the message to be processed based on the reminder method.

2. The method according to claim 1, wherein: The determining, based on the context information of the message to be processed, the emotion category of the message to be processed and the model fields required by the preset message model includes: Using a feature extractor to extract target features from the message to be processed and the context information; Inputting the target feature into a sentiment classification head to obtain the sentiment category of the message to be processed; and The target feature is input into the key information classification header to obtain the model field.

3. The method according to claim 2, wherein: The feature extractor includes an encoder, a multi-head self-attention mechanism network module and a hidden layer; The extracting target features from the message to be processed and the context information by using a feature extractor includes: The encoder is used to encode the message to be processed and the context information to obtain encoding features; the encoding features are used to model the semantics and context environment of the message to be processed and the context information; Using the multi-head self-attention mechanism network module to process the encoded features to parse out the core semantic features in the message to be processed and the context information; The core semantic features are processed using a hidden layer to obtain the target features.

4. The method according to claim 2, wherein: The key information classification head includes a fully connected layer and a discriminator; The step of inputting the target feature into a key information classification header to obtain the model field comprises: Processing the target feature based on the fully connected layer to obtain a feature to be discriminated; Inputting the discriminant feature into the discriminator to obtain the weight of each word in the message to be processed, and determining the urgency of the message to be processed; Identifying key fields from the message to be processed based on the weights of each word segment; The model fields include the urgency level and the key fields.

5. The method according to claim 1, further comprising creating the preset message model based on the following method: Obtain historical group chat messages and the target object's interaction records with the historical group chat messages; Inputting the historical group chat messages and the interaction records into a large model, so that the large model analyzes and summarizes a plurality of description statements; Obtaining group chat messages corresponding to the plurality of description statements in the historical group chat messages; Based on the group chat messages corresponding to the multiple description sentences, the preset message model is constructed.

6. The method according to claim 1, further comprising creating the preset message model based on the following method: Mining keywords from the chat group based on keyword mining technology; The preset message model is constructed based on the keywords.

7. The method according to claim 1, further comprising creating the preset message model based on the following method: In response to configuration information input by the target object on a message setting page; Based on the configuration information, the preset message model is constructed.

8. The method according to claim 1, wherein: The determining of the reminder method of the message to be processed includes: Determining a reminder level of the pending message; In the reminder method set, the reminder method corresponding to the reminder level is filtered.

9. The method according to any one of claims 1 to 8, further comprising: Based on the feedback of the target object on the message to be processed, the preset message model set is updated.

10. A group chat message processing device, comprising: The acquisition module is used to obtain the pending messages in the chat group; A first determination module, configured to determine the emotion category of the message to be processed and the model fields required by the preset message model based on the context information of the message to be processed; A matching module, used to match the emotion category and model field of the message to be processed with a preset message model set of a target object; A second determination module is used to determine a reminder mode of the message to be processed when any preset message model in the preset message model set is matched; The reminder module is used to remind the pending message based on the reminder method.

11. The device according to claim 10, wherein: The first determining module includes: An extraction unit, configured to extract target features from the message to be processed and the context information using a feature extractor; A first classification unit is used to input the target feature into a sentiment classification head to obtain the sentiment category of the message to be processed; and The second classification unit is used to input the target feature into the key information classification header to obtain the model field.

12. The device according to claim 11, wherein The feature extractor includes an encoder, a multi-head self-attention mechanism network module and a hidden layer; the extraction unit is specifically used for: The encoder is used to encode the message to be processed and the context information to obtain encoding features; the encoding features are used to model the semantics and context environment of the message to be processed and the context information; Using the multi-head self-attention mechanism network module to process the encoded features to parse out the core semantic features in the message to be processed and the context information; The core semantic features are processed using a hidden layer to obtain the target features.

13. The device according to claim 11, wherein: The key information classification head includes a fully connected layer and a discriminator; the second classification unit is specifically used for: Processing the target feature based on the fully connected layer to obtain a feature to be discriminated; Inputting the discriminant feature into the discriminator to obtain the weight of each word in the message to be processed, and determining the urgency of the message to be processed; Identifying key fields from the message to be processed based on the weights of each word segment; The model fields include the urgency level and the key fields.

14. The apparatus according to claim 10, further comprising a first building module, configured to: Obtain historical group chat messages and the target object's interaction records with the historical group chat messages; Inputting the historical group chat messages and the interaction records into a large model, so that the large model analyzes and summarizes a plurality of description statements; Obtaining group chat messages corresponding to the plurality of description statements in the historical group chat messages; Based on the group chat messages corresponding to the multiple description sentences, the preset message model is constructed.

15. The apparatus according to claim 10, further comprising a second building block, configured to: Mining keywords from the chat group based on keyword mining technology; The preset message model is constructed based on the keywords.

16. The apparatus according to claim 10, further comprising a third building block, configured to: In response to configuration information input by the target object on a message setting page; Based on the configuration information, the preset message model is constructed.

17. The device according to claim 10, wherein: The second determining module is used to: Determining a reminder level of the pending message; In the reminder method set, the reminder method corresponding to the reminder level is filtered.

18. The apparatus according to any one of claims 10 to 17, further comprising a feedback module, configured to: Based on the feedback of the target object on the message to be processed, the preset message model set is updated.

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

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

21. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images