Information processing method and apparatus, device, and computer readable medium

Abstract

The application relates to an information processing method, device and equipment and a computer readable medium. The application uses an AI large model to identify malicious comment data, feeds the comment environment before and after filtering to the AI large model, summarizes the problems discussed before and after filtering, the comment tendency before and after filtering, summarizes the comments with similar opinions, and reflects the real public opinion after filtering to the user, thereby solving the technical problem that the user is easily trapped in the information cocoon room due to the inability to identify problem comments.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to an information processing method, apparatus, device, and computer-readable medium. Background Art

[0002] With the rapid development of the internet, people's online social interactions and consumption patterns have undergone tremendous changes. From the relatively closed, independent internet forums of the early days, to the more comprehensive Baidu Tieba and Douban groups, and now to the inclusive and open Weibo, people's online social interactions have become increasingly open and inclusive. Online e-commerce has also rapidly grown alongside the internet, and a significant portion of consumption has shifted from offline to online. However, open online communities have their upsides. This open environment has also led to the emergence of a large number of comment-posting bots, hired "trolls," and "retroactive" comments. Furthermore, the platforms' personalized recommendation algorithms restrict users to only what the platforms want them to see, easily trapping them in information cocoons.

[0003] At present, although there are plug-ins in related technologies that force users to browse their comments and dynamics, they are still unable to identify robot comments, "water army" comments or "reverse-dressing" comments, resulting in the comment information seen by users may not be the most authentic comment information, and users are easily trapped in an information cocoon.

[0004] There is currently no effective solution to the problem that users are easily trapped in information cocoons due to the inability to identify problematic comments. Summary of the Invention

[0005] The present application provides an information processing method, apparatus, device and computer-readable medium to solve the technical problem that users are easily trapped in information cocoons due to the inability to identify problematic comments.

[0006] According to one aspect of an embodiment of the present application, the present application provides an information processing method, including: obtaining initial comment data and initial comment environment information of a target topic, wherein the target topic is an issue discussed jointly by multiple accounts in a social network; inputting the initial comment data and the initial comment environment information into a target big model, so as to summarize the initial comment background of the target topic based on the initial comment environment information through the target big model, and identifying malicious comment data in the initial comment data based on the initial comment background, wherein the malicious comment data is comment data that guides public opinion away from the target topic; filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information; inputting the target comment data and the target comment environment information into the target big model, so as to summarize the real comment background of the target topic based on the target comment environment information through the target big model, and classifying and summarizing the target comment data based on the real comment background.

[0007] Optionally, obtaining the initial comment data and initial comment environment information of the target topic includes: obtaining the comment data to be identified under the target topic as the initial comment data; determining the account that issued the initial comment data as the account to be identified; obtaining the account information and historical dynamic information of the account to be identified, wherein the account information includes the registration time, account level and account name of the account; and determining the account information and the historical dynamic information of each account to be identified as the initial comment environment information.

[0008] Optionally, before inputting the initial comment data and the initial comment environment information into the target big model, the method also includes obtaining the target big model in the following manner: obtaining malicious comment samples and account data samples corresponding to the malicious comment samples; analyzing and organizing the account data samples to obtain the account basic information, historical dynamic information and the problem information being discussed when each malicious comment sample is published; inputting the malicious comment samples, the account basic information, the historical dynamic information and the problem information into the initial multi-category classification model for deep learning, so that the initial multi-category classification model performs weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the account basic information, the historical dynamic information and the problem information to learn to identify various types of malicious comment samples; performing regression verification on the trained initial multi-category classification model, and when the verification result meets the target condition, determining the initial multi-category classification model as the target big model.

[0009] Optionally, the obtaining of malicious comment samples and account data samples corresponding to the malicious comment samples includes at least one of the following: obtaining comments whose account level is less than a first threshold and whose creation time is less than a second threshold, and whose comment content similarity is greater than a third threshold as robot comment samples, and obtaining the account information and historical dynamic information of the robot comment samples as the account data samples corresponding to the robot comment samples, wherein the malicious comment samples include the robot comment samples; obtaining comments whose number of comments appearing at the same time is greater than a fourth threshold and whose comment content similarity is greater than a fifth threshold as water army comment samples, and obtaining the account information and historical dynamic information of the water army comment samples as the account data samples corresponding to the water army comment samples, wherein the malicious comment samples include the water army comment samples; obtaining comments whose similarity between the comment content and the account user portrait is less than a sixth threshold and whose number of comments is greater than a seventh threshold as anti-play comment samples, and obtaining the account information and historical dynamic information of the anti-play comment samples as the account data samples corresponding to the anti-play comment samples, wherein the malicious comment samples include the anti-play comment samples.

[0010] Optionally, the initial multi-category classification model performs weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the account basic information, the historical dynamic information and the problem information to learn to identify various types of malicious comment samples, including at least one of the following: when it is determined that the account corresponding to the malicious comment sample has an account level less than the first threshold, a creation time less than the second threshold, and an account name regularity less than a preset degree threshold, the classification weight of the robot comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the robot comment; when it is determined that multiple malicious comment samples appear at the same time and the similarity of the comment content is greater than the fifth threshold, the classification weight of the water army comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the water army comment; when it is determined that multiple malicious comment samples appear at the same time and the similarity of the comment content to the account user portrait is less than the sixth threshold, the classification weight of the anti-playing comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the anti-playing comment.

[0011] Optionally, after classifying and summarizing the target comment data based on the real comment background, the method also includes: determining the proportion of the malicious comment data and the target comment data in the initial comment data, and drawing a diagram based on their respective proportion data; displaying the diagram, marking the malicious comment data, and comparing the marked malicious comment data with the target comment data to reflect the real public opinion of the target topic while prompting the user to identify the malicious comment data.

[0012] Optionally, the method further includes: opening a data interface to a target browser plug-in, so that the target browser plug-in processes comment information in the social network through the data interface, wherein the target browser plug-in is embedded in the browser for front-end display.

[0013] According to another aspect of an embodiment of the present application, the present application provides an information processing device, including: an acquisition module, used to acquire initial comment data and initial comment environment information of a target topic, wherein the target topic is an issue discussed jointly by multiple accounts in a social network; an identification module, used to input the initial comment data and the initial comment environment information into a target big model, so as to summarize the initial comment background of the target topic based on the initial comment environment information through the target big model, and identify malicious comment data in the initial comment data based on the initial comment background, wherein the malicious comment data is comment data that guides public opinion away from the target topic; a filtering module, used to filter out the malicious comment data in the initial comment data to obtain target comment data, and filter out the malicious comment environment information corresponding to the malicious comment data in the initial comment environment information to obtain target comment environment information; a summarizing module, used to input the target comment data and the target comment environment information into the target big model, so as to summarize the real comment background of the target topic based on the target comment environment information through the target big model, and classify and summarize the target comment data based on the real comment background.

[0014] According to another aspect of an embodiment of the present application, the present application provides an electronic device, including a memory, a processor, a communication interface and a communication bus, wherein the memory stores a computer program that can be run on the processor, the memory and the processor communicate through the communication bus and the communication interface, and the steps of the above method are implemented when the processor executes the computer program.

[0015] According to another aspect of the embodiments of the present application, the present application also provides a computer-readable medium having a non-volatile program code executable by a processor, where the program code enables the processor to execute the above method.

[0016] The above technical solution provided by the embodiment of the present application has the following advantages compared with the related art:

[0017] The present application provides an information processing method, including: obtaining initial comment data and initial comment environment information of a target topic, wherein the target topic is an issue discussed jointly by multiple accounts in a social network; inputting the initial comment data and the initial comment environment information into a target big model, so as to summarize the initial comment background of the target topic based on the initial comment environment information through the target big model, and identifying malicious comment data in the initial comment data based on the initial comment background, wherein the malicious comment data is comment data that guides public opinion away from the target topic; filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information; inputting the target comment data and the target comment environment information into the target big model, so as to summarize the real comment background of the target topic based on the target comment environment information through the target big model, and classifying and summarizing the target comment data based on the real comment background. This application uses an AI big model to identify malicious comment data, feeds the comment environment before and after filtering to the AI ​​big model, summarizes and outputs the issues discussed before and after filtering, as well as the comment tendencies before and after filtering, summarizes comments with similar opinions, and reflects the real public opinion after filtering to users, solving the technical problem that users are easily trapped in information cocoons due to the inability to identify problematic comments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 A schematic diagram of a hardware environment for an optional information processing method provided according to an embodiment of the present application;

[0021] Figure 2 A flowchart of an optional information processing method provided according to an embodiment of the present application;

[0022] Figure 3 This is a block diagram of an optional information processing device provided according to an embodiment of the present application;

[0023] Figure 4 A schematic diagram of an optional electronic device structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the accompanying drawings for the embodiments of the present application to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments are a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0025] In the following description, the suffixes such as “module”, “part” or “unit” used to represent elements are only for the convenience of the description of the present application, and have no specific meaning in itself. Therefore, “module” and “part” can be used interchangeably.

[0026] In order to solve the problems mentioned in the background, according to an aspect of the embodiments of the present application, an embodiment of an information processing method is provided.

[0027] Optionally, in the embodiments of the present application, the above-mentioned information processing method can be applied to a hardware environment composed of a terminal 101 and a server 103 as shown in the figure. As shown in the figure, the server 103 is connected with the terminal 101 through a network, which can be used to provide services for the terminal or the client installed on the terminal, and a database 105 can be set on the server or independently of the server, which is used to provide data storage services for the server 103. The above-mentioned network includes but is not limited to a wide area network, a metropolitan area network or a local area network, and the terminal 101 includes but is not limited to a PC, a mobile phone, a tablet computer and the like. Figure 1 Figure 1

[0028] The information processing method in the embodiments of the present application can be executed by the server 103, and can also be executed by the server 103 and the terminal 101 together, as shown in the figure. The method can include the following steps: Figure 2

[0029] Step S202, obtaining initial comment data and initial comment environment information of a target topic, wherein the target topic is a question discussed by multiple accounts in a social network;

[0030] Step S204, inputting the initial comment data and the initial comment environment information into a target large model, so as to summarize an initial comment background of the target topic based on the initial comment environment information through the target large model, and identify malicious comment data in the initial comment data based on the initial comment background, wherein the malicious comment data is comment data leading public opinion to deviate from the target topic;

[0031] ​​​Step S206, filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information;

[0032] Step S208 inputs the target comment data and the target comment environment information into the target large model to summarize the real comment background of the target topic based on the target comment environment information through the target large model, and classifies and summarizes the target comment data based on the real comment background.

[0033] In the embodiments of the present application, the target topic refers to a specific problem or topic discussed by multiple accounts in a social network. The initial comment data refers to all original comment contents published by users on the social network for the target topic. The initial comment environment information refers to environmental factors associated with the initial comment data, including comment publishing time, location, publisher information, emotional tendency, etc., as well as the association relationship between comments (such as reply, like, etc.). The target large model is a pre-trained model based on deep learning technology, which can process natural language text and associated information, perform comment background summarization, malicious comment identification, comment classification and summarization, etc.

[0034] With the popularity of social networks, users tend to express their own opinions and comments on the platform. However, the existence of malicious comments can interfere with normal discussions, mislead public opinion, and even cause adverse effects on individuals or society. To this end, the present application uses AI large model to identify malicious comment data, feeds the comment environment before and after filtering to the AI large model, summarizes and outputs the problems discussed before and after filtering, and the comment tendency before and after filtering, reminds users to pay attention to distinguishing problem comments, and reflects the real public opinion after filtering to users. Specifically:

[0035] In an embodiment of the present application, initial comment data and initial comment environment information about the target topic are collected from social networks through web crawlers or other data acquisition technologies. The initial data is input into the target big model, and the big model will first summarize the initial comment background of the target topic based on the initial comment environment information. Then, the big model uses the malicious comment recognition ability learned in advance, combined with the initial comment background, to identify the malicious comment data in the initial comment data. The identified malicious comment data is removed from the initial comment data to obtain the target comment data that does not contain malicious comments. At the same time, the information corresponding to the malicious comment data is removed from the initial comment environment information to obtain the target comment environment information. The target comment data and the target comment environment information are input into the target big model again. The big model summarizes the real comment background of the target topic based on more real and effective target comment environment information. Based on the real comment background, the big model classifies the target comment data (such as positive, negative, neutral, etc.) and generates a summary report that summarizes the core viewpoints, trends and emotional attitudes discussed in the target topic.

[0036] In the embodiment of the present application, in addition to text-type comment data, multimodal comment data (such as pictures, videos, audio, etc.) can be collected by extending the multimodal data interface to conduct a more comprehensive analysis and understanding of the user's comment intentions and emotions.

[0037] In the embodiments of the present application, when processing user data, the relevant laws and regulations on privacy protection and data security are strictly observed to ensure the security and legal use of user data, and sensitive information is desensitized to prevent user privacy leakage.

[0038] This application uses an AI big model to identify malicious comment data, feeds the comment environment before and after filtering to the AI ​​big model, summarizes and outputs the issues discussed before and after filtering, as well as the comment tendencies before and after filtering, summarizes comments with similar opinions, and reflects the real public opinion after filtering to users, solving the technical problem that users are easily trapped in information cocoons due to the inability to identify problematic comments.

[0039] In an optional embodiment, the step of obtaining the initial review data and initial review environment information of the target topic includes:

[0040] Step 1: Obtain the comment data to be identified under the target topic as the initial comment data;

[0041] Step 2: Determine the account that sent the initial comment data as the account to be identified;

[0042] Step 3: Obtain the account information and historical dynamic information of the account to be identified, wherein the account information includes the account registration time, account level and account name;

[0043] Step 4: Determine the account information and the historical dynamic information of each account to be identified as the initial comment environment information.

[0044] In an embodiment of the present application, all comment data under the target topic is captured from the social networking platform through a web crawler or other data acquisition technology as the comment data to be identified. Based on the publisher information of the comment data to be identified, the account of the issuer of each comment, i.e., the account to be identified, is determined. Basic information such as the registration time, account level, and account name of the account to be identified is obtained from the account management system of the social networking platform. Through the API interface or data crawling technology, the historical content, likes, comments, shares, and other interactions posted by the account to be identified on the social networking platform are obtained. The account information and historical dynamic information of each account to be identified are integrated and associated to form initial comment environment information for subsequent data processing and analysis.

[0045] This application obtains the comment data to be identified and directly analyzes the discussion content under the target topic, obtains the account information and historical dynamic information of the account to be identified as the initial comment environment information, so as to fully understand the identity characteristics, behavior patterns, possible positions and biases of the commentator, thereby improving the accuracy and reliability of malicious comment identification.

[0046] In an optional embodiment, before inputting the initial review data and the initial review environment information into the target macro model, the method further includes obtaining the target macro model in the following manner:

[0047] Step 1: Obtain malicious comment samples and account data samples corresponding to the malicious comment samples;

[0048] Step 2: Analyze and organize the account data samples to obtain basic account information, historical dynamic information, and information about the issues being discussed when each malicious comment sample was posted.

[0049] Step 3: Inputting the malicious comment samples, the basic account information, the historical dynamic information, and the problem information into an initial multi-category classification model for deep learning, so that the initial multi-category classification model performs a weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the basic account information, the historical dynamic information, and the problem information to learn to identify various types of the malicious comment samples;

[0050] Step 4: Perform regression verification on the trained initial multi-category classification model, and when the verification result meets the target condition, determine the initial multi-category classification model as the target large model.

[0051] In the embodiments of this application, malicious comment samples refer to confirmed malicious comment data on social networks, such as bot comments, "troll" comments, or "reverse-acting" comments. Account data samples are information about the social network accounts corresponding to the malicious comment samples, including basic account information and historical dynamic information. The initial multi-class classification model is a deep learning model used for multi-class classification.

[0052] In an embodiment of the present application, the training process of the target large model is specifically as follows: a certain number of malicious comment samples and corresponding account data samples are collected. The collected account data samples are analyzed and sorted in detail to extract the basic information of the account (such as registration time, account level, etc.), historical dynamic information, and information about the issues being discussed when each malicious comment is posted. The malicious comment samples, basic account information, historical dynamic information, and problem information are input as input data into the initial multi-category classification model. Through deep learning technology, the model learns how to classify malicious comment samples based on the input data, and optimizes the accuracy of the classification results through weighted scoring analysis. An independent verification data set is used to perform regression verification on the trained initial multi-category classification model. If the verification result meets the preset target conditions (such as the accuracy rate, recall rate, etc. reach a certain level), the model is determined as the final target large model; otherwise, it is necessary to adjust the model parameters or collect samples again for training.

[0053] By collecting and analyzing real malicious comment samples and their corresponding account data samples, this application can ensure that the model learns the characteristics and patterns of malicious comments, improving the accuracy and reliability of recognition. At the same time, through multi-category classification and weighted scoring analysis, the model can better distinguish different types of malicious comments and adapt to recognition needs in different environments and scenarios.

[0054] In an optional embodiment, obtaining the malicious comment sample and the account data sample corresponding to the malicious comment sample includes at least one of the following:

[0055] Obtaining comments whose account level is less than a first threshold and whose creation time is less than a second threshold, and whose comment content similarity is greater than a third threshold as robot comment samples, and obtaining account information and historical dynamic information of the robot comment samples as the account data samples corresponding to the robot comment samples, wherein the malicious comment samples include the robot comment samples;

[0056] Obtaining comments whose number of comments appearing in the same time period is greater than a fourth threshold and whose comment content similarity is greater than a fifth threshold as water army comment samples, and obtaining account information and historical dynamic information of the water army comment samples as the account data samples corresponding to the water army comment samples, wherein the malicious comment samples include the water army comment samples;

[0057] Obtain comments whose similarity between the comment content and the account user portrait is less than the sixth threshold and whose number of comments is greater than the seventh threshold as anti-stringing comment samples, and obtain the account information and historical dynamic information of the anti-stringing comment samples as the account data samples corresponding to the anti-stringing comment samples, wherein the malicious comment samples include the anti-stringing comment samples.

[0058] In this embodiment, newly created, low-level accounts that post a large number of similar comments within a short period of time are considered bot comments. The specific process for obtaining bot comment samples is as follows: a first threshold and a second threshold are set to filter out accounts with a low account level and creation time, respectively. The comments of these accounts are checked and similarity is calculated. When the similarity exceeds a third threshold, these comments are considered bot comment samples. The account information and historical dynamic information of these comment samples are collected as account data samples.

[0059] In the embodiments of this application, water army comments typically post a large number of similar comments within a short period of time to create public opinion or mislead the public. Comments with this characteristic are considered water army comments. The specific process for obtaining water army comment samples is as follows: monitor the comment platform, search for accounts with a number of comments greater than a fourth threshold appearing in the same period, examine the comment content of these accounts, calculate similarity, and when the similarity exceeds a fifth threshold, identify these comments as water army comment samples. The account information and historical dynamic information of these comment samples are collected as account data samples.

[0060] In the embodiments of the present application, comments that are inconsistent with the account's regular behavior or user profile in order to intentionally cause controversy or confuse the public can be considered as anti-dressing comments. The specific process for obtaining anti-dressing comment samples is as follows: analyze the account's user profile, including regular behavior, interest preferences, etc., check the account's comment content, calculate the similarity with the user profile, and when the similarity is less than the sixth threshold and the number of comments is greater than the seventh threshold, these comments are regarded as anti-dressing comment samples, and the account information and historical dynamic information of these comment samples are collected as account data samples.

[0061] In the embodiment of the present application, the specific settings of the first to seventh thresholds can be reasonably set according to specific circumstances and data characteristics.

[0062] In an optional embodiment, the initial multi-category classification model performs a weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the basic account information, the historical dynamic information, and the question information to learn to identify each type of the malicious comment samples, including at least one of the following:

[0063] When it is determined that the account corresponding to the malicious comment sample has any of an account level less than the first threshold, a creation duration less than the second threshold, or a regularity degree of an account name less than a preset degree threshold, the classification weight of the robot comment is increased by a set step to improve the score of classifying the malicious comment sample as the robot comment.

[0064] When it is determined that the multiple malicious comment samples occur at the same time and have a comment content similarity greater than the fifth threshold, the classification weight of the water army comment is increased by a set step to improve the score of classifying the malicious comment sample as the water army comment.

[0065] When it is determined that the multiple malicious comment samples occur at the same time and have a comment content similarity with the account user portrait less than the sixth threshold, the classification weight of the anti-string comment is increased by a set step to improve the score of classifying the malicious comment sample as the anti-string comment.

[0066] In the embodiments of the present application, different types of malicious comments often have different features and patterns when training a malicious comment identification model. Therefore, the present application performs weighted analysis on different classifications. Specifically, in the training process, the input malicious comment sample and its corresponding account data sample are subjected to feature extraction and analysis. The features include account level, creation duration, regularity degree of account name, historical dynamics, etc. According to the results of feature analysis, the classification weights of different types of malicious comments in the initial multi-class classification model are adjusted. For robot comments, when the account level is less than the first threshold, the creation duration is less than the second threshold, or the regularity degree of the account name is less than the preset degree threshold (such as the account name is a random combination of numbers and letters), the classification weight of the robot comment is increased. For water army comments, when multiple malicious comment samples occur at the same time and have a comment content similarity greater than the fifth threshold, the classification weight of the water army comment is increased. For anti-string comments, when multiple malicious comment samples occur at the same time and have a comment content similarity with the account user portrait less than the sixth threshold, the classification weight of the anti-string comment is increased. The higher the weight, the higher the score of the corresponding category. Finally, the model identifies the malicious comment data to be classified as the type with the highest score.

[0067] By adjusting the classification weight, the model can pay more attention to the features and patterns of different types of malicious comments, thereby improving the identification accuracy of different types of malicious comments. This helps to improve the generalization ability and robustness of the model, making it better adapt to malicious comment identification tasks in different environments and scenarios.

[0068] In an optional embodiment, after classifying and summarizing the target comment data based on the real comment background, the method further comprises:

[0069] Step 1, determine the proportion of the malicious comment data and the target comment data in the initial comment data, and draw a graph according to the respective proportion data;

[0070] Step 2, display the graph and mark the malicious comment data, and display the marked malicious comment data and the target comment data for comparison, so as to reflect the real public opinion of the target topic while prompting the user to identify the malicious comment data.

[0071] In the embodiments of the present application, after the target comment data is classified and summarized by the target large model, the relationship between the malicious comment data and the target comment data can be further processed and analyzed, and these data can be displayed in a visual manner, so that the user can more intuitively understand the real public opinion of the target topic and identify the malicious comment. Specifically, first, collect initial comment data, and use a multi-class classification model to classify and summarize the target comment data, then count the number or frequency of malicious comment data and target comment data, and calculate their proportion in the initial comment data. According to the proportion data, select a suitable graph type (such as pie chart, column chart, etc.), and use a drawing tool or library (such as Matplotlib, Plotly, etc.) to draw a graph. In addition, the malicious comment data can be directly marked on the social network using colors, texts or other visual elements, so that the user can identify. The marked malicious comment data and the target comment data are displayed for comparison, which can be side-by-side graphs, or can be distinguished by adjusting the color, size, etc. of the visual elements.

[0072] The present application displays the comments before and after filtering in a visual manner, so that the user can more intuitively understand the real public opinion of the target topic and identify the malicious comment.

[0073] In an optional embodiment, the method further comprises:

[0074] A data interface is opened to the target browser plug-in, so that the target browser plug-in processes the comment information in the social network through the data interface, wherein the target browser plug-in is embedded in the browser for front-end display.

[0075] In the embodiments of the present application, first, the standard and specification of the data interface need to be defined, including the name, parameters, return value, etc. of the interface. These interfaces should be able to support the operations of obtaining, processing and analyzing the comment information in the social network. According to the defined data interface specification, the corresponding backend service or API is developed for processing the request from the browser plug-in and returning the corresponding data or result. The plug-in developer develops the target browser plug-in according to the demand. The plug-in needs to include the function of communicating with the data interface in order to obtain data or send requests from the backend service. The developed browser plug-in is embedded into the browser so that the target browser plug-in can process the comment information in the social network through the data interface.

[0076] Through the browser plug-in, the user can directly operate and process the comment information in the social network in the browser environment without jumping to other applications or platforms, greatly improving the convenience and continuity of user experience. The browser plug-in, as an extension of the browser function, can provide more customized services for users. Through the open data interface, plug-in developers can develop various functions based on these data interfaces, such as comment filtering, sentiment analysis, content recommendation, etc. The browser plug-in usually runs on the user's local device and has stronger data processing capability. Through the data interface, part of the data processing task can be transferred to the plug-in end to reduce the burden of the server end.

[0077] The present application provides an information processing method, including: obtaining initial comment data and initial comment environment information of a target topic, wherein the target topic is an issue discussed jointly by multiple accounts in a social network; inputting the initial comment data and the initial comment environment information into a target big model, so as to summarize the initial comment background of the target topic based on the initial comment environment information through the target big model, and identifying malicious comment data in the initial comment data based on the initial comment background, wherein the malicious comment data is comment data that guides public opinion away from the target topic; filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information; inputting the target comment data and the target comment environment information into the target big model, so as to summarize the real comment background of the target topic based on the target comment environment information through the target big model, and classifying and summarizing the target comment data based on the real comment background. This application uses an AI big model to identify malicious comment data, feeds the comment environment before and after filtering to the AI ​​big model, summarizes and outputs the issues discussed before and after filtering, as well as the comment tendencies before and after filtering, summarizes comments with similar opinions, and reflects the real public opinion after filtering to users, solving the technical problem that users are easily trapped in information cocoons due to the inability to identify problematic comments.

[0078] According to another aspect of the embodiment of the present application, Figure 3 As shown, an information processing device is provided, comprising:

[0079] An acquisition module 301 is configured to acquire initial comment data and initial comment context information of a target topic, wherein the target topic is a topic discussed by multiple accounts in a social network;

[0080] Identification module 303 is configured to input the initial comment data and the initial comment context information into a target macro model, so as to summarize the initial comment context of the target topic based on the initial comment context information through the target macro model, and identify malicious comment data in the initial comment data based on the initial comment context, wherein the malicious comment data is comment data that leads public opinion away from the target topic;

[0081] The filtering module 305 is configured to filter out the malicious comment data from the initial comment data to obtain target comment data, and to filter out the malicious comment context information corresponding to the malicious comment data from the initial comment context information to obtain target comment context information;

[0082] The summarizing module 307 is used to input the target review data and the target review environment information into the target big model, so as to summarize the real review background of the target topic based on the target review environment information through the target big model, and classify and summarize the target review data based on the real review background.

[0083] It should be noted that the acquisition module 301 in this embodiment can be used to execute step S202 in the embodiment of the present application, the identification module 303 in this embodiment can be used to execute step S204 in the embodiment of the present application, the filtering module 305 in this embodiment can be used to execute step S206 in the embodiment of the present application, and the summary module 307 in this embodiment can be used to execute step S208 in the embodiment of the present application.

[0084] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. Figure 1 In the hardware environment shown, it can be implemented by software or by hardware.

[0085] Optionally, the acquisition module is specifically used to: obtain the comment data to be identified under the target topic as the initial comment data; determine the account that issued the initial comment data as the account to be identified; obtain the account information and historical dynamic information of the account to be identified, wherein the account information includes the account registration time, account level and account name; determine the account information and the historical dynamic information of each account to be identified as the initial comment environment information.

[0086] Optionally, the information system processing device also includes a model training module, which is specifically used to: obtain malicious comment samples and account data samples corresponding to the malicious comment samples; analyze and organize the account data samples to obtain the account basic information, historical dynamic information and the problem information being discussed when each malicious comment sample is published; input the malicious comment samples, the account basic information, the historical dynamic information and the problem information into the initial multi-category classification model for deep learning, so that the initial multi-category classification model performs weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the account basic information, the historical dynamic information and the problem information to learn to identify various types of malicious comment samples; perform regression verification on the trained initial multi-category classification model, and when the verification result meets the target condition, determine the initial multi-category classification model as the target large model.

[0087] Optionally, the model training module is further configured to: obtain comments with an account level less than a first threshold value, a creation time length less than a second threshold value, and a comment content similarity greater than a third threshold value as robot comment samples, and obtain account information and historical dynamic information of the robot comment samples as the account data samples corresponding to the robot comment samples, wherein the malicious comment samples include the robot comment samples; obtain comments with a comment quantity greater than a fourth threshold value and a comment content similarity greater than a fifth threshold value at the same time as water army comment samples, and obtain account information and historical dynamic information of the water army comment samples as the account data samples corresponding to the water army comment samples, wherein the malicious comment samples include the water army comment samples; and obtain comments with a comment content similarity to an account user portrait less than a sixth threshold value and a comment quantity greater than a seventh threshold value as anti-string comment samples, and obtain account information and historical dynamic information of the anti-string comment samples as the account data samples corresponding to the anti-string comment samples, wherein the malicious comment samples include the anti-string comment samples.

[0088] Optionally, the model training module is further configured to: when determining that the account corresponding to the malicious comment sample has any of an account level less than the first threshold value, a creation time length less than the second threshold value, and a regularity degree of an account name less than a preset degree threshold value, increase a classification weight of a robot comment by a set step to improve a score of classifying the malicious comment sample as the robot comment; when determining that a plurality of the malicious comment samples appear at the same time and have a comment content similarity greater than the fifth threshold value, increase a classification weight of a water army comment by a set step to improve a score of classifying the malicious comment sample as the water army comment; and when determining that a plurality of the malicious comment samples appear at the same time and have a comment content similarity to an account user portrait less than the sixth threshold value, increase a classification weight of an anti-string comment by a set step to improve a score of classifying the malicious comment sample as the anti-string comment.

[0089] Optionally, the information processing apparatus further includes a data visualization module, which is specifically configured to: determine proportions of the malicious comment data and the target comment data in the initial comment data, and draw a graph according to the respective proportions; display the graph, mark the malicious comment data, and compare and display the marked malicious comment data and the target comment data, so as to reflect real public opinions of the target topic and prompt a user to identify the malicious comment data.

[0090] Optionally, the information processing device further includes an interface module, which is specifically used to: open a data interface to a target browser plug-in so that the target browser plug-in processes comment information in the social network through the data interface, wherein the target browser plug-in is embedded in the browser for front-end display.

[0091] According to another aspect of the embodiment of the present application, the present application provides an electronic device, such as Figure 4 As shown, it includes a memory 401, a processor 403, a communication interface 405 and a communication bus 407. The memory 401 stores a computer program that can be run on the processor 403. The memory 401 and the processor 403 communicate through the communication interface 405 and the communication bus 407. When the processor 403 executes the computer program, the steps of the above method are implemented.

[0092] The memory and processor in the electronic device communicate via a communication bus and a communication interface. The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus may be divided into an address bus, a data bus, a control bus, and the like.

[0093] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0094] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0095] According to another aspect of the embodiments of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of any of the above embodiments.

[0096] Optionally, in an embodiment of the present application, the computer-readable medium is configured to store program codes for the processor to execute the following steps:

[0097] Obtaining initial comment data and initial comment context information for a target topic, wherein the target topic is a topic discussed by multiple accounts in a social network;

[0098] Inputting the initial comment data and the initial comment context information into a target macro model, so as to summarize the initial comment context of the target topic based on the initial comment context information through the target macro model, and identifying malicious comment data in the initial comment data based on the initial comment context, wherein the malicious comment data is comment data that leads public opinion away from the target topic;

[0099] Filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information;

[0100] The target review data and the target review environment information are input into the target big model, so as to summarize the real review background of the target topic based on the target review environment information through the target big model, and classify and summarize the target review data based on the real review background.

[0101] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0102] When implementing the embodiments of the present application, reference may be made to the above embodiments, which have corresponding technical effects.

[0103] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or a combination thereof.

[0104] For software implementation, the technology described herein can be implemented by a unit that performs the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0105] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0109] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0110] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of software products, and the computer software products are stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, and various program codes that can be stored in the medium. It should be noted that in this paper, terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices including a series of elements not only include those elements, but also include other elements not explicitly listed, or include elements inherent to such processes, methods, articles or devices. Without more limitations, the elements defined by the statement "including a" do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0111] The above is only a specific embodiment of the present application, so that those skilled in the art can understand or implement the present application. Various modifications of these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features applied herein.

Claims

1. An information processing method, characterized in that: include: Obtaining initial comment data and initial comment context information for a target topic, wherein the target topic is a topic discussed by multiple accounts in a social network; Inputting the initial comment data and the initial comment context information into a target macro model, so as to summarize the initial comment context of the target topic based on the initial comment context information through the target macro model, and identifying malicious comment data in the initial comment data based on the initial comment context, wherein the malicious comment data is comment data that leads public opinion away from the target topic; Filtering out the malicious comment data from the initial comment data to obtain target comment data, and filtering out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information; Inputting the target review data and the target review environment information into the target macro model, so as to summarize the real review background of the target subject based on the target review environment information through the target macro model, and classifying and summarizing the target review data based on the real review background; Before inputting the initial review data and the initial review environment information into the target macro model, the method further includes obtaining the target macro model in the following manner: Obtain malicious comment samples and account data samples corresponding to the malicious comment samples; Analyze and organize the account data samples to obtain basic account information, historical dynamic information, and information about the issues being discussed when each malicious comment sample was posted. Inputting the malicious comment samples, the basic account information, the historical dynamic information, and the problem information into an initial multi-category classification model for deep learning, so that the initial multi-category classification model performs a weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the basic account information, the historical dynamic information, and the problem information to learn to identify various types of the malicious comment samples; The trained initial multi-category classification model is subjected to regression verification, and when the verification result meets the target condition, the initial multi-category classification model is determined as the target large model.

2. The method according to claim 1, characterized in that The acquisition of the initial review data and initial review environment information of the target topic includes: Acquire the comment data to be identified under the target topic as the initial comment data; Determine the account that sent the initial comment data as the account to be identified; Obtaining account information and historical dynamic information of the account to be identified, wherein the account information includes account registration time, account level, and account name; The account information and the historical dynamic information of each account to be identified are determined as the initial comment environment information.

3. The method according to claim 1, characterized in that The acquiring of malicious comment samples and account data samples corresponding to the malicious comment samples includes at least one of the following: Obtaining comments whose account level is less than a first threshold and whose creation time is less than a second threshold, and whose comment content similarity is greater than a third threshold as robot comment samples, and obtaining account information and historical dynamic information of the robot comment samples as the account data samples corresponding to the robot comment samples, wherein the malicious comment samples include the robot comment samples; Obtaining comments whose number of comments appearing in the same time period is greater than a fourth threshold and whose comment content similarity is greater than a fifth threshold as water army comment samples, and obtaining account information and historical dynamic information of the water army comment samples as the account data samples corresponding to the water army comment samples, wherein the malicious comment samples include the water army comment samples; Obtain comments whose similarity between the comment content and the account user portrait is less than the sixth threshold and whose number of comments is greater than the seventh threshold as anti-stringing comment samples, and obtain the account information and historical dynamic information of the anti-stringing comment samples as the account data samples corresponding to the anti-stringing comment samples, wherein the malicious comment samples include the anti-stringing comment samples.

4. The method according to claim 3, characterized in that The enabling the initial multi-category classification model to perform weighted scoring analysis on the classification results of the malicious comment samples based on the malicious comment samples, the basic account information, the historical dynamic information, and the question information to learn to identify each type of the malicious comment samples includes at least one of the following: When it is determined that the account corresponding to the malicious comment sample has any of the following conditions: the account level is less than the first threshold, the creation time is less than the second threshold, and the regularity of the account name is less than the preset threshold, the classification weight of the robot comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the robot comment; When it is determined that a plurality of the malicious comment samples appear at the same time and the similarity of the comment content is greater than the fifth threshold, the classification weight of the water army comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the water army comment; When it is determined that multiple malicious comment samples appear at the same time and the similarity between the comment content and the account user portrait is less than the sixth threshold, the classification weight of the anti-character comment is increased according to the set step size to improve the score of classifying the malicious comment sample as the anti-character comment.

5. The method according to claim 1, wherein After classifying and summarizing the target review data based on the real review background, the method further includes: Determine the proportion of the malicious comment data and the target comment data in the initial comment data, and draw a graph based on the respective proportion data; The diagram is displayed, and the malicious comment data is marked, and the marked malicious comment data is compared with the target comment data for display, so as to reflect the real public opinion of the target topic while prompting the user to identify the malicious comment data.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: A data interface is opened to a target browser plug-in, so that the target browser plug-in processes comment information in a social network through the data interface, wherein the target browser plug-in is embedded in a browser for front-end display.

7. An information processing device for implementing the information processing method according to any one of claims 1 to 6, characterized in that: include: an acquisition module, configured to acquire initial comment data and initial comment context information of a target topic, wherein the target topic is a topic discussed by multiple accounts in a social network; an identification module, configured to input the initial comment data and the initial comment context information into a target macro model, so as to summarize the initial comment context of the target topic based on the initial comment context information through the target macro model, and to identify malicious comment data in the initial comment data based on the initial comment context, wherein the malicious comment data is comment data that leads public opinion away from the target topic; a filtering module configured to filter out the malicious comment data from the initial comment data to obtain target comment data, and to filter out the malicious comment environment information corresponding to the malicious comment data from the initial comment environment information to obtain target comment environment information; A summarizing module is used to input the target review data and the target review environment information into the target big model, so as to summarize the real review background of the target topic based on the target review environment information through the target big model, and classify and summarize the target review data based on the real review background.

8. An electronic device comprising a memory, a processor, a communication interface, and a communication bus, wherein the memory stores a computer program that can be run on the processor, and the memory and the processor communicate via the communication bus and the communication interface, characterized in that: When the processor executes the computer program, the information processing method according to any one of claims 1 to 6 is implemented.

9. A computer-readable medium having a non-volatile program code executable by a processor, characterized in that The program code enables the processor to execute the information processing method according to any one of claims 1 to 6.

Images