User forum viewpoint extraction and analysis method and system and storage medium

Abstract

The invention discloses a user forum viewpoint extraction and analysis method and system, and the method comprises the steps: crawling the data of each post from a forum through a web crawler technology; guiding the large model to execute the following steps by utilizing a CoT technology: identifying concern points of each user in post data, summarizing the concern points into different topics and de-duplicating, evaluating the occurrence frequency and importance of each topic, and outputting each topic and description thereof; based on the topics output by the large model and the descriptions of the topics, the large model is used for classifying the posts in the forum to be analyzed, and if the posts do not belong to the topics output by the large model, the posts are listed as other categories; and extracting viewpoints of the users in the posts corresponding to the topics by using the large model, obtaining a plurality of representative viewpoints in one topic, and carrying out summarization and analysis according to the viewpoints. According to the method, the forum viewpoints can be efficiently extracted and analyzed.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and specifically relates to a method, system, electronic device and storage medium for extracting and analyzing user forum opinions. Background Art

[0002] Online discussion forums provide critical data for understanding broad real-world community concerns. However, traditional methods for analyzing these data, such as training models for specific forums and topics, are difficult to scale and require large amounts of training data. These limitations have prompted researchers to look for more efficient and scalable analysis methods.

[0003] Chinese patent CN116821457A-An intelligent consultation and opinion processing system based on a multimodal large model, proposes an intelligent consultation and opinion processing system based on a multimodal large model; the processing system determines a main opinion topic involved in the user input by performing context analysis on the user input, and guides the user to conduct a neutral discussion. However, the system is a consultation question-and-answer system that requires continuous interaction with users for analysis. It is not an automated forum data analysis system and cannot automatically realize online forum data and opinion analysis.

[0004] Chinese Patent CN117910575A - A method for monitoring and evaluating network opinion risk events. A method for monitoring and evaluating network opinion risk events is proposed, which mainly includes: a network opinion data resource library, a network opinion event extraction model based on a large model, and a network opinion event risk value calculation. The system is mainly used to assess the overall risk of opinion events, but it can only monitor and extract events, and does not conduct in-depth analysis of the text data of online forums; the system focuses more on monitoring broad opinion events and cannot conduct in-depth analysis of specific groups or forums.

[0005] Chinese Patent CN117350301A - A method for monitoring opinion events based on a multimodal fusion model. A method for monitoring opinion events based on a multimodal fusion model is proposed. The method segments and filters the images and audio in the multimodal document data, retains the images and audio with the highest similarity to the text, combines them with the selected text blocks, and inputs them into the multimodal model for analysis. The system focuses more on monitoring and identifying opinion events, and cannot extract the main opinions of each forum user in the event; the system does not conduct in-depth analysis and evaluation of opinion events. Summary of the invention

[0006] In view of the above problems, the present invention provides a method, system and storage medium for extracting and analyzing user forum opinions, aiming to efficiently extract and analyze forum opinions.

[0007] According to a first aspect of an embodiment of the present disclosure, a method for extracting and analyzing user forum opinions is provided, the method comprising the following steps:

[0008] Use web crawler technology to crawl data from each post on the forum;

[0009] The CoT technology is used to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description;

[0010] Based on the topics and descriptions output by the large model, the large model is used to classify the posts in the forum to be analyzed. If the posts do not belong to the topics output by the large model, they are classified as other categories;

[0011] The large model is used to extract the opinions of users in the posts corresponding to each topic. Several representative opinions are obtained for each topic, and a summary and analysis are performed based on the opinions.

[0012] In some embodiments, the post data includes a post title, a post body, a timestamp, posting user information, and a comment list.

[0013] In some embodiments, when the large model outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the large model itself. If it is determined to be the knowledge of the large model itself, the corresponding output content will be deleted.

[0014] In some embodiments, the large model is a large language model LLM, and Prompt is used to guide the large language model LLM to output each topic and its description.

[0015] According to a second aspect of an embodiment of the present disclosure, a user forum opinion extraction and analysis system is provided, the system comprising:

[0016] The data acquisition module is used to crawl the data of each post from the forum using web crawler technology;

[0017] The topic extraction module is used to use the CoT technology to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description;

[0018] A post classification module is used to classify posts in the analysis forum using the big model based on the topics and descriptions output by the big model. If a post does not belong to the topics output by the big model, it is classified as other categories.

[0019] The opinion evaluation module is used to use the large model to extract the opinions of users in the posts corresponding to each topic, obtain several representative opinions for one topic, and summarize and analyze them based on the opinions.

[0020] In some embodiments, the post data in the data acquisition module includes post title, post body, timestamp, posting user information and comment list.

[0021] In some embodiments, when the large model in the topic extraction module outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the large model itself. If it is determined to be the knowledge of the large model itself, the corresponding output content will be deleted.

[0022] In some embodiments, the large model of the system is a large language model LLM, and Prompt is used to guide the large language model LLM to output each topic and its description.

[0023] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned user forum opinion extraction and analysis method when executing the program.

[0024] According to a fourth aspect of an embodiment of the present disclosure, there is provided a non-temporary computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the above-mentioned user forum opinion extraction and analysis method.

[0025] The disclosed embodiments provide a method, system and storage medium for extracting and analyzing user forum opinions. Aiming at online forum text data analysis scenarios, a method for extracting and analyzing opinions based on a large language model is proposed, which includes an innovative prompt word method and evaluation strategy for efficiently extracting and analyzing forum opinions. This method first crawls forum data from online forums based on specific topics, uses a large model to extract the main concerns of forum users, then classifies each concern according to the topic, analyzes the most representative concerns of each subject, and finally summarizes. The entire opinion extraction and analysis process is automatically completed by the large model, and no human participation is required. Its beneficial effects include:

[0026] Forum data analysis method based on big model: Use big model as analysis tool to automatically process and analyze text data of online forums, classify user topics, and extract the main points of each topic. This method has multi-domain scalability and is compatible with different data sources. It can be applied to medical, financial, education and other fields. At the same time, this method can also process data from different online forums and social media platforms.

[0027] Innovative prompt word method and multi-step analysis process: A novel prompt word method was developed, which guides LLM to perform tasks such as topic extraction, classification and evaluation more accurately through a series of customized prompts. A multi-step analysis process was constructed to ensure the comprehensiveness and depth of the analysis results.

[0028] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present invention and, together with the description, serve to explain the principles of the present invention.

[0030] Figure 1 1 is a flow chart of a method for extracting and analyzing user forum opinions in an embodiment of the present invention;

[0031] Figure 2 It is a logic block diagram of a method for extracting and analyzing user forum opinions in an embodiment of the present invention;

[0032] Figure 3 1 is a schematic diagram of the structure of a user forum opinion extraction and analysis system in an embodiment of the present invention;

[0033] Figure 4 It is a schematic diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0035] It should be mentioned before discussing the exemplary embodiments in more detail that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the steps as sequential processes, many of the steps therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of the steps can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the accompanying drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0036] Some terminology descriptions based on the embodiments of the present invention:

[0037] LLM: Large Language Model;

[0038] Prompt: The prompt is the input or prompt given to the artificial intelligence model. This can be a piece of text, a question, an instruction, or any form of input to guide the model to produce a specific output or response.

[0039] CoT: Chain of Thought, a method of showing the model's thinking process when solving a problem, breaking the problem into a series of steps, and then solving each step step by step to finally get the answer.

[0040] The present invention provides the following embodiments for a method, system and storage medium for extracting and analyzing user forum opinions:

[0041] Example 1 is used to illustrate the method for extracting and analyzing user forum opinions, see Figure 1 , which is a flow chart of the method for extracting and analyzing user forum opinions. The specific steps include:

[0042] S1. Use web crawler technology to crawl data from each post on the forum;

[0043] Specifically, the post data includes post title, post body, timestamp, posting user information and comment list.

[0044] S2. Use CoT technology to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description;

[0045] Specifically, when the big model outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the big model itself. If it is determined to be the knowledge of the big model itself, the corresponding output content will be deleted.

[0046] S3. Based on the topics and descriptions output by the large model, the large model is used to classify the posts in the forum to be analyzed. If the posts do not belong to the topics output by the large model, they are classified as other categories.

[0047] S4. Use the big model to extract the opinions of users in the posts corresponding to each topic, obtain several representative opinions for one topic, and summarize and analyze them based on the opinions.

[0048] Specifically, the large model is a large language model LLM, and Prompt is used to guide the large language model LLM to output each topic and its description.

[0049] In the specific implementation process, Figure 2 As shown:

[0050] Data preparation stage: Use web crawler technology to crawl data from forums on specific topics. The crawled data content includes but is not limited to post titles, post text, timestamps, posting user information, and comment lists.

[0051] Topic extraction stage: Use the CoT technology of the big model to tell the big model the data format that needs to be analyzed, and specify the following analysis steps:

[0052] The first step is to identify the focus of each user in the post;

[0053] Step 2: Group the focus into different topics and remove duplicates;

[0054] Step 3: Evaluate the frequency and importance of each topic.

[0055] Step 4: Output the description of each topic;

[0056] Step 5. Make sure the output content comes from the post itself, not the big model’s own knowledge.

[0057] Post classification stage: Based on the topics and descriptions summarized in the previous step, use the big model to classify each post, and ensure that each post is classified either into the topic summarized in the previous step or into the "other" category.

[0058] Opinion evaluation stage: For posts on each topic, use the big model to extract the main opinions of users in the relevant posts, and select the most representative opinions on each topic. Finally, summarize and analyze based on these opinions.

[0059] Among them, Figure 2 As shown, the embodiment also provides the following specific examples:

[0060] Take the analysis of Rideshare (the topic of online ride-hailing drivers in the Reddit forum) forum data as an example: First, use a crawler to collect forum data. The format of each data is (post title, post content, post time, post ID, comment list). After the topic extraction stage, the content in the forum is divided into several main topics, each of which contains: (title, short description, representative post ID). In this scenario, a possible topic is "Online ride-hailing platforms need to enhance transparency and explainability: due to the lack of information and explanations about artificial intelligence and algorithmic features, drivers need this information in their work. These issues include opaque trip details, unclear peak pricing boundaries, bonus and task clarity issues, lack of salary segmentation, platform commission rate and clarity on how to calculate prices and bonuses, and the impact of ratings on indicators." All topics and descriptions are added to the prompt of the big model, requiring the big model to classify all posts in the forum, and the classification results are added to the forum dataset. At this time, the format of each data in the dataset is (post title, post content, post time, post ID, comment list, topic). Next, for each topic post, use the big model to summarize the 5 most representative opinions and give the corresponding post IDs. For the topic in the above example, the possible opinions are:

[0061] (1) Drivers face uncertain and volatile income due to changes in peak and off-peak pricing (representative comment: “When ride-hailing platforms increase their promotional efforts, they manipulate and lower peak pricing”);

[0062] (2) There are challenges in understanding and adapting to algorithmic changes that affect trip allocation and revenue;

[0063] (3) The complexity and unpredictability of qualifying for incentive programs such as tasks and bonuses;

[0064] (4) expressed concern about the inadequate payment for the time and distance invested in long pick-ups and waits;

[0065] (5) An increase in the number of drivers results in a decrease in available trips and lower revenue.

[0066] Among them, Prompt is used to guide the large language model LLM to output each topic and its description, specifically including:

[0067] Topic Extraction: Analyze a set of JSON objects, each representing a post from the Rideshare forum. For each JSON object, the following information will be found:

[0068] Submission Title: The title of the post.

[0069] Submission Body: The main content or message of your post.

[0070] Timestamp: The date and time the post was submitted.

[0071] ID: A unique identifier for all posts in the dataset.

[0072] Comments: A list of all comments on this post.

[0073] Generate a list of the most frequent and influential concerns due to AI and platform algorithms discussed by posters in the input context.

[0074] Step 1: Identify concerns (points of concern) mentioned in the submission body, title, and / or comments related to drivers’ lack of understanding or available information about the platform’s algorithms, including but not limited to fees, routes, incentive programs, driver preferences, etc.

[0075] Step 2: Group similar concerns and avoid redundancy. For example, multiple mentions of the cost calculation problem should be grouped into the same concern.

[0076] Step 3: Select the citation that best represents each concern from the grouped concerns. Make sure the citation clearly states the specific concerns that arise due to AI and platform capabilities.

[0077] Step 4: Evaluate which concerns are mentioned most frequently and have the greatest impact on drivers.

[0078] Step 5: Create a list of these concerns in JSON format. Each entry should contain (with the following specific field names):

[0079] "title": the title of the focus

[0080] "description": a short description

[0081] "quote": the selected representative quotation

[0082] Step 6: Make sure the final list is concise, precise, and contains only the concerns found in the input context, without generalizing based on pre-trained or external knowledge.

[0083] Step 7: Group any similar concerns to avoid redundancy.

[0084] Post Classification: The task is to analyze a list of issues regarding transparency on ride-hailing platforms. Please evaluate each post and classify it into one of the five categories below. Please assign a letter to each listed issue, corresponding to its most appropriate category, based on the order of the original list (keeping the order of the issues unchanged).

[0085] category:

[0086] A. Need for greater transparency and explainability: Due to the lack of information and explanation about the platform algorithms, drivers cannot understand their work, which leads to problems. These problems include opaque trip details, unclear boundaries of peak hours, problems with bonus and task clarity, lack of salary details, platform commissions, and how prices and bonuses are calculated, and the impact of ratings on indicators.

[0087] B. Need for greater predictability and worker autonomy: Issues arising from reduced worker autonomy and less predictable working conditions due to significant changes in algorithmic features. These issues include income instability, misleading destination screening, peak-time pricing and matching that varies widely across drivers and locations, and deceptive trip quotes.

[0088] C. Need for better safety and more time: Issues arising from safety risks and time pressures due to algorithmic features. These issues include dangerous multi-tasking, compromised route safety, unfinished tasks, and acceptance rate issues, highlighting the need for improved safety and time management in application design.

[0089] D. Need to ensure fairness and non-discrimination: Issues related to algorithmic wage discrimination. These issues include unequal pay for similar work, income below the minimum wage, and the impact of driver and passenger demographics on income, all of which are caused by opaque algorithmic features.

[0090] E. Other issues: issues that do not fall into any of the above categories.

[0091] Output format: Please present the analysis results in dictionary format, with the question number as the key and the category as the value, and keep the order of the questions. For example: {1:A,2:B,3:C}

[0092] Note: Make sure each post is classified into only one of the above categories and there is one category for each question in the input, e.g. for 400 posts there should be 400 entries in the dictionary. Do not generate any other text.

[0093] Opinion Assessment: This data contains a set of questions from ride-hailing drivers obtained from discussions on the Rideshare forum.

[0094] The categories of these questions are as follows:

[0095] <Insert category description from Category Prompt>

[0096] Find the top 5 most common topics of concern that are most relevant to ride-hailing drivers. Make sure each question is clearly distinct from the others on the list. Do not repeat the same question in the list. If similar questions exist, group them together and find another question to ask.

[0097] Output Format: Provides output in ranking format:

[0098] {concern_rank:,concern_title:string,concern_description:a string of 10-20 words}

[0099] Finally, each post in the forum dataset is classified into one of the following 6 categories:

[0100] AE: <5 most common topics of interest summarized in the previous step>

[0101] F: Other topics

[0102] Output format: Please present the analysis results in dictionary format, with the post number as the key and the focus topic as the value, and keep the order of the questions. For example: {1:A,2:B,3:C}.

[0103] Note: Make sure each post is classified into only one of the above categories and that there is one category for each post in the input, e.g. for 400 posts there should be 400 entries in the dictionary. Do not generate any other text.

[0104] Another embodiment is used to illustrate the user forum opinion extraction and analysis system, see Figure 3 , the system 300 comprises:

[0105] The data acquisition module 310 is used to crawl the data of each post from the forum using the web crawler technology;

[0106] The topic extraction module 320 is used to use the CoT technology to guide the large model to perform the following steps: identifying the focus of each user in the post data, summarizing the focus into different topics and removing duplicates, evaluating the frequency and importance of each topic, and outputting each topic and its description;

[0107] A post classification module 330 is used to classify each post in the forum to be analyzed based on each topic and its description output by the large model, and if a post does not belong to each topic output by the large model, it is classified as other categories;

[0108] The opinion evaluation module 340 is used to extract the opinions of users in the posts corresponding to each topic using the large model, obtain several representative opinions for one topic, and summarize and analyze based on the opinions.

[0109] The post data in the data acquisition module 310 includes the post title, post body, timestamp, posting user information and comment list.

[0110] When the large model in the topic extraction module 320 outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the large model itself. If it is determined to be the knowledge of the large model itself, the corresponding output content will be deleted.

[0111] The large model of the system 300 is a large language model LLM, and Prompt is used to guide the large language model LLM to output various topics and their descriptions.

[0112] In addition to the upper module, the system 300 may also include other components. However, since these components are irrelevant to the content of the embodiment of the present disclosure, their illustration and description are omitted here.

[0113] The other specific working processes of the user forum opinion extraction and analysis system 300 refer to the description of the above-mentioned user forum opinion extraction and analysis method embodiment, which will not be repeated here.

[0114] Another embodiment is used to illustrate that the system of the present invention can also be used with the help of Figure 4 The architecture of the computing device shown is implemented. Figure 4 The architecture of the computing device is shown. Figure 4 As shown, a computer system 410, a system bus 430, one or more CPUs 440, an input / output 420, a memory 450, etc. The memory 450 can store various data or files used for computer processing and / or communication and program instructions executed by the CPU including the user forum opinion extraction and analysis method of the embodiment. Figure 4 The architecture shown is only exemplary and can be adjusted according to actual needs when implementing different devices. Figure 4 One or more components in the system. The memory 450, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the user forum opinion extraction and analysis method in the embodiment of the present invention (for example, the data acquisition module 310, the topic extraction module 320, the post classification module 330 and the opinion evaluation module 340 in the user forum opinion extraction and analysis system 300). One or more CPUs 440 execute various functional applications and data processing of the system of the present invention by running the software programs, instructions and modules stored in the memory 450, that is, to implement the above-mentioned user forum opinion extraction and analysis method, which includes the following steps:

[0115] Use web crawler technology to crawl data from each post on the forum;

[0116] The CoT technology is used to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description;

[0117] Based on the topics and descriptions output by the large model, the large model is used to classify the posts in the forum to be analyzed. If the posts do not belong to the topics output by the large model, they are classified as other categories;

[0118] The large model is used to extract the opinions of users in the posts corresponding to each topic. Several representative opinions are obtained for each topic, and a summary and analysis are performed based on the opinions.

[0119] Of course, the processor of the server provided in the embodiment of the present invention is not limited to executing the method operations described above, but can also execute related operations in the user forum opinion extraction and analysis method provided in any embodiment of the present invention.

[0120] The memory 450 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 450 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 450 may further include a memory remotely arranged relative to one or more CPUs 440, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0121] The input / output 420 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The input / output 420 may also include a display device such as a display screen.

[0122] The embodiment of the present invention also provides a non-temporary computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by the processor, the user forum opinion extraction and analysis method described in the above embodiment is implemented. The computer-readable storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, 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 above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0123] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0124] The program code contained on the storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0125] In addition, other specific working processes of a non-temporary computer-readable storage medium refer to the description of the above-mentioned user forum opinion extraction and analysis method embodiment and will not be repeated here.

[0126] Combining the technical solutions provided by the above embodiments, a method and system for extracting and analyzing user forum opinions, aimed at online forum text data analysis scenarios, proposes an opinion extraction and analysis method based on a large language model, including an innovative prompt word method and evaluation strategy for efficiently extracting and analyzing forum opinions. This method first crawls forum data from online forums based on specific topics, uses a large model to extract the main concerns of forum users, then classifies each concern according to the topic, analyzes the most representative concerns of each subject, and finally summarizes. The entire opinion extraction and analysis process is automatically completed by the large model and does not require human participation. Its beneficial effects include:

[0127] Forum data analysis method based on big model: Use big model as analysis tool to automatically process and analyze text data of online forums, classify user topics, and extract the main points of each topic. This method has multi-domain scalability and is compatible with different data sources. It can be applied to medical, financial, education and other fields. At the same time, this method can also process data from different online forums and social media platforms.

[0128] Innovative prompt word method and multi-step analysis process: A novel prompt word method was developed, which guides LLM to perform tasks such as topic extraction, classification and evaluation more accurately through a series of customized prompts. A multi-step analysis process was constructed to ensure the comprehensiveness and depth of the analysis results.

[0129] In this document, the terms "comprises," "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, such that a step or method that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such step or method.

[0130] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for extracting and analyzing user forum opinions, characterized in that: The method comprises the following steps: Use web crawler technology to crawl data from each post on the forum; The CoT technology is used to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description; Based on the topics and descriptions output by the large model, the large model is used to classify the posts in the forum to be analyzed. If the posts do not belong to the topics output by the large model, they are classified as other categories; The large model is used to extract the opinions of users in the posts corresponding to each topic. Several representative opinions are obtained for each topic, and a summary and analysis are performed based on the opinions.

2. The method for extracting and analyzing user forum opinions according to claim 1, characterized in that: The post data includes post title, post body, timestamp, posting user information and comment list.

3. The method for extracting and analyzing user forum opinions according to claim 1, characterized in that: When the large model outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the large model itself. If it is determined to be the knowledge of the large model itself, the corresponding output content will be deleted.

4. The method for extracting and analyzing user forum opinions according to claim 1, characterized in that: The large model is a large language model LLM, and Prompt is used to guide the large language model LLM to output various topics and descriptions thereof.

5. A user forum opinion extraction and analysis system, characterized in that: The system comprises: The data acquisition module is used to crawl the data of each post from the forum using web crawler technology; The topic extraction module is used to use the CoT technology to guide the large model to perform the following steps: identify the focus of each user in the post data, summarize the focus into different topics and remove duplicates, evaluate the frequency and importance of each topic, and output each topic and its description; A post classification module is used to classify posts in the analysis forum using the big model based on the topics and descriptions output by the big model. If a post does not belong to the topics output by the big model, it is classified as other categories. The opinion evaluation module is used to use the large model to extract the opinions of users in the posts corresponding to each topic, obtain several representative opinions for one topic, and summarize and analyze them based on the opinions.

6. The user forum opinion extraction and analysis system according to claim 5, characterized in that: The post data in the data acquisition module includes post title, post body, timestamp, posting user information and comment list.

7. The user forum opinion extraction and analysis system according to claim 5, characterized in that: When the large model in the topic extraction module outputs each topic and its description, it is also necessary to determine whether the output content comes from the post or the knowledge of the large model itself. If it is determined to be the knowledge of the large model itself, the corresponding output content will be deleted.

8. The user forum opinion extraction and analysis system according to claim 5, characterized in that: The large model of the system is a large language model LLM, and Prompt is used to guide the large language model LLM to output various topics and their descriptions.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the user forum opinion extraction and analysis method according to any one of claims 1 to 4 are implemented.

10. A non-transitory computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the user forum opinion extraction and analysis method as described in any one of claims 1 to 4 are implemented.

Images