A product public opinion report generation method, device and system

Abstract

The embodiment of the application provides a product public opinion report generation method, device and system, and relates to the technical field of artificial intelligence. The method comprises the following steps: collecting and preprocessing evaluation data to obtain multiple groups of evaluation data classified based on data types; performing data sampling on each group of evaluation data based on the importance sampling principle to obtain a sampling data group of each data type; performing public opinion analysis on each sampling data group respectively to obtain initial intermediate information; and generating a public opinion report based on the initial intermediate information. The method extracts effective data from massive resources based on the importance sampling principle, and then generates a public opinion report by using the effective data, so that the processing efficiency can be effectively improved, the same type of data overload can be avoided, the consumption of computing resource is reduced, and the report generation efficiency and report quality are improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, and system for generating product public opinion reports. Background Technology

[0002] When conducting public opinion analysis on car reviews, reports are usually generated directly from massive resources. However, the effective information extracted from these resources may contain too much data of the same type, leading to biased analysis results and affecting the quality of the report. In addition, too much data of the same type can also affect the timeliness of the report. Summary of the Invention

[0003] The purpose of this application is to provide a product public opinion report generation method, device and system. Based on the importance sampling principle, effective data is extracted from massive resources and then used to generate public opinion reports. This can effectively improve processing efficiency, avoid overload of similar data, reduce computing power consumption, and improve report generation efficiency and report quality.

[0004] In a first aspect, this application provides a method for generating a product public opinion report. The method includes: collecting and preprocessing evaluation data to obtain multiple sets of evaluation data based on data type classification; sampling each set of evaluation data based on the importance sampling principle to obtain a sample data set for each data type; performing public opinion analysis on each sample data set to obtain initial intermediate information; and generating a public opinion report based on the initial intermediate information.

[0005] In the technical solution of this application embodiment, the method samples from massive data based on the importance sampling principle, extracts effective data, and then performs public opinion analysis based on the extracted sampled data group to generate a public opinion report. Compared with directly generating public opinion reports using massive data, this method extracts effective data based on the importance sampling principle, which can effectively improve processing efficiency, avoid overload of similar data, reduce computing resource consumption, improve report generation efficiency and quality, and improve the timeliness of report generation. It solves the problem that existing methods may have too much similar data when directly generating public opinion reports using massive data, which may lead to one-sided analysis results and affect the quality of public opinion reports. At the same time, too much similar data affects the timeliness of report generation.

[0006] In some embodiments, evaluation data is collected and preprocessed to obtain multiple sets of evaluation data based on data type classification, including: collecting evaluation data based on user-inputted requirements; cleaning the collected evaluation data and marking it by data type and source; classifying the evaluation data based on the data type to obtain multiple sets of evaluation data, wherein the data type includes text, image, chart, and video. Obtaining multiple sets of evaluation data based on data type facilitates subsequent data sampling based on grouping through data classification.

[0007] In some embodiments, data sampling is performed on each set of evaluation data based on the importance sampling principle to obtain a sample data set for each data type. This includes: calculating the importance weight of each public opinion information sample in each set of evaluation data based on the importance sampling principle; sorting each set of evaluation data in descending order based on the importance weight; and obtaining the top M public opinion information samples as the sample data set based on the sorting result, where M is a set value. Data sampling based on the importance sampling principle can obtain effective data for report generation, which can effectively improve data analysis efficiency and reduce computing resource consumption compared to directly using massive amounts of data.

[0008] In some embodiments, the importance weight of each public opinion information sample in each set of evaluation data is calculated based on the importance sampling principle, including: calculating the target distribution of a single public opinion information sample based on pre-set evaluation indicators and corresponding weights. The proposal distribution is determined based on the most readily available and weighted evaluation metrics. The importance weight of each individual public opinion information sample is determined based on the target distribution and the proposal distribution. : Importance weights are determined based on the target distribution and the proposed distribution, so that the data can be filtered using these weights to obtain effective data.

[0009] In some embodiments, the target distribution of a single public opinion information sample is represented as follows: ;in, Indicates the first i Sample of public opinion information Target distribution These are samples of a single piece of public opinion information. The calculation results of the corresponding evaluation indicators, including viewership, interaction rate, professionalism, authority, and timeliness. The weights corresponding to the evaluation indicators are respectively and The target distribution of a single public opinion information sample is calculated using evaluation indicators and corresponding weights, so that the importance weight can be calculated subsequently using the target distribution.

[0010] In some embodiments, the sampled data set includes a text data set, an image data set, a chart data set, and a video data set. The step of performing public opinion analysis on each sampled data set to obtain initial intermediate information includes: sending the text data set to a text analysis model to obtain sentiment intensity and evaluation focus information; sending the image data set to an image analysis model to obtain product part attention and usage scenario data; sending the chart data set to a chart analysis model to obtain market sales attention and driving range attention; and sending the video data set to a video analysis model to obtain key testing steps, quantitative data on test results, and sentiment trend data. Performing public opinion analysis on the sampled data sets using corresponding models and employing parallel collaborative processing can improve processing efficiency.

[0011] Secondly, this application provides a product public opinion report generation device, which includes: a data acquisition module for collecting and preprocessing evaluation data to obtain multiple sets of evaluation data based on data type classification; a sampling module for sampling each set of evaluation data based on the importance sampling principle to obtain sampled data sets for each data type; a public opinion analysis module for performing public opinion analysis on each set of sampled data to obtain initial intermediate information; and a report generation module for generating a public opinion report based on the initial intermediate information. Utilizing the importance sampling principle to sample data, obtain effective data, and then perform public opinion analysis can effectively improve processing efficiency, avoid overload of similar data, reduce computing resource consumption, and improve report generation efficiency and report quality.

[0012] Thirdly, this application provides a product public opinion report generation system, which includes: the product public opinion report generation device described in the second aspect and a coordination and command module, wherein: the coordination and command module is used to decompose the public opinion report generation task into: data collection and preprocessing sub-tasks, data sampling sub-tasks, public opinion analysis sub-tasks, and report generation sub-tasks based on user-input requirements, and sequentially distribute them to each processing module in the product public opinion report generation device for processing, and perform task scheduling during the processing. By decomposing the public opinion report generation task into multiple sub-tasks based on user-input requirements, a multi-module parallel processing strategy can be used under the unified management of the coordination and command module to efficiently generate more comprehensive, in-depth, and reliable product public opinion analysis reports, such as those for automobile evaluations.

[0013] Fourthly, this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the above-described product public opinion report generation method.

[0014] Fifthly, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the above-described product public opinion report generation method.

[0015] Sixthly, this application provides a computer program product, which includes a computer program that, when run by a processor, executes the above-described product public opinion report generation method. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a product public opinion report generation method provided in this application embodiment; Figure 2 A flowchart of data acquisition and preprocessing provided for embodiments of this application; Figure 3 A data sampling flowchart provided for embodiments of this application; Figure 4 A flowchart illustrating the calculation of importance weights provided in the embodiments of this application; Figure 5 Structural block diagram of the product public opinion report generation device provided in the embodiments of this application; Figure 6 This is a structural block diagram of a product public opinion report generation system provided in an embodiment of this application.

[0018] icon: 101-Data Acquisition Module; 102-Sampling Module; 103-Public Opinion Analysis Module; 104-Report Generation Module; 105-Coordination and Command Module. Detailed Implementation

[0019] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] When generating product public opinion reports, such as car review public opinion reports, it is usually necessary to obtain massive amounts of data from various car websites. However, directly using the obtained massive amounts of data for public opinion analysis is not only inefficient but may also lead to an overload of similar data. The results of public opinion analysis may be biased, and it will also affect the quality and timeliness of the public opinion report.

[0022] To address the aforementioned technical issues, this application provides a method for generating product public opinion reports. This method is based on the principle of importance sampling, extracts effective data from massive resources, and then uses the effective data to generate public opinion reports. This can effectively improve processing efficiency, avoid overload of similar data, reduce the consumption of computing resources, and improve report generation efficiency, report quality, and timeliness.

[0023] Please refer to Figure 1 , Figure 1 A flowchart of a product public opinion report generation method provided in this application embodiment, the method includes the following steps: S110: Collect and preprocess the evaluation data to obtain multiple sets of evaluation data based on data type classification; S120: Based on the importance sampling principle, data sampling is performed on each set of evaluation data to obtain sample data sets for each data type; S130: Perform public opinion analysis on each group of sampled data to obtain initial intermediate information; S140: Generate a public opinion report based on initial intermediate information.

[0024] The data types here include text, images, charts, and videos. Four types of evaluation data can be obtained based on the data type, allowing for data sampling for each set of evaluation data based on the importance sampling principle. Initial intermediate information includes sentiment tendency and sentiment intensity obtained from text analysis, and user-interested car-related scenarios obtained from image analysis.

[0025] This method classifies the massive amounts of collected data and uses the importance sampling principle to sample the data, thereby obtaining effective data for public opinion analysis. Compared with directly generating public opinion reports from massive amounts of data, this method can effectively improve processing efficiency, avoid overload of similar data, reduce the consumption of computing resources, and improve report generation efficiency, report quality, and timeliness.

[0026] Please refer to Figure 2 , Figure 2 This is a flowchart of data acquisition and preprocessing. In some embodiments, evaluation data is acquired and preprocessed to obtain multiple sets of evaluation data based on data type classification, including: S111: Collect evaluation data based on user-inputted requirements; S112: Clean the collected evaluation data and mark its data type and source; S113: Classify the evaluation data based on data type to obtain multiple sets of evaluation data, where data types include text, images, charts, and videos.

[0027] User-inputted requirements, also known as public opinion report requirements, can take various forms, including but not limited to user-defined report outlines, plain text requirements, and user-uploaded public opinion report templates. User-inputted requirements are generally vague and general, therefore requiring a deep understanding of user intent and accurate identification of core concerns. This can be achieved by extracting keywords and expanding the semantics of the requirements, transforming user needs into precise and executable system tasks.

[0028] For the collection of evaluation data, based on the target vehicle model, time range, amount of information, data type, and source obtained from the demand analysis, raw data related to public opinion needs is crawled from different channels in parallel and stored in the raw data cache. Information channels include, but are not limited to: mainstream media and news websites, social media, online forums, and video websites. The information obtained includes, but is not limited to: evaluation articles, evaluation videos and their subtitles, scoring data, images, and charts.

[0029] The collected evaluation data is preprocessed: the raw data is read from the raw data cache and cleaned by deduplication, ad removal and removal of irrelevant information. The raw data is then content-identified, and the data is labeled with its data type and source. The data is then divided into four groups according to type: "text, image, chart and video".

[0030] The main function of data preprocessing is to classify data, standardize data formats, and provide clean and consistent data for subsequent analysis.

[0031] Please refer to Figure 3 , Figure 3 This is a flowchart of the data sampling process. In some embodiments, data is sampled for each set of evaluation data based on the importance sampling principle to obtain sampled data sets for each data type, including: S121: Calculate the importance weight of each public opinion information sample in each set of evaluation data based on the importance sampling principle; S122: Sort each set of evaluation data in descending order based on importance weight; S123: Obtain the top M public opinion information samples as a sampling data group based on the sorting results, where M is a set value.

[0032] The importance sampling method used in this application involves sorting the data in the text, image, chart, and video categories according to their "importance weight" from high to low, and then extracting a user-specified number of data points from each group. The principle behind importance sampling is as follows: In the calculation function In target distribution When the expected value is: ; directly from the target distribution Sampling is difficult in medium-sized regions, so a proposed distribution that is easier to sample is introduced. ,So: ; Therefore, from Sampling Then, the expected value is estimated using a weighted average: ; in, This refers to the importance weight.

[0033] Based on the aforementioned importance sampling principle, the importance weight of each public opinion information sample in each set of evaluation data is calculated. This allows for the extraction of data with higher importance weights from each set of evaluation data as sampling data groups, ensuring that all types of data are covered. This effectively avoids the problem of data overload of the same type and prevents the one-sidedness of public opinion analysis results. Furthermore, given the limitations of computing resources and the timeliness of report generation, selecting a certain number of representative data from massive datasets for data sampling can improve analysis efficiency, focus on key value data to improve analysis quality, reduce computing resource consumption, and improve the timeliness of report generation.

[0034] Please refer to Figure 4 , Figure 4 This is a flowchart illustrating the calculation of importance weights. In some embodiments, the importance weight of each public opinion information sample in each set of evaluation data is calculated based on the importance sampling principle, including: S124: Calculate the target distribution of a single public opinion information sample based on pre-set evaluation indicators and corresponding weights. ; S125: Determine the proposal distribution based on the most readily available and weighted evaluation metrics. ; S126: Determine the importance weight of a single public opinion information sample based on the target distribution and the proposal distribution. : .

[0035] First, it is necessary to define the key indicators for product evaluation, such as viewership, interaction rate, professionalism, authority, and timeliness. Evaluation indicators should be determined from multiple dimensions to make the obtained target distribution closer to the current evaluation sentiment.

[0036] The proposal distribution is determined by the most readily available and highest-weighted evaluation metric, resulting in a highly accurate distribution. This ultimately leads to highly accurate importance weights, enabling precise sampling to obtain effective data.

[0037] In some embodiments, the target distribution of a single public opinion information sample is represented as follows: ; in, Indicates the first i Sample of public opinion information Target distribution These are samples of a single piece of public opinion information. The corresponding evaluation metrics are calculated, including viewership, interaction rate, professionalism, authority, and timeliness. The weights corresponding to the evaluation metrics are respectively and .

[0038] The table below shows the specific evaluation indicators, including viewership, interaction rate, professionalism, authority, and timeliness.

[0039] Table 1 Evaluation Indicators and Weights

[0040] Among them, viewership is the number of page views, which can be accurately obtained and therefore has a high weight; the interaction rate is the proportion of comments, reposts and likes to the number of views.

[0041] Professionalism refers to the level of expertise in product reviews, which can be quantified using ratings. Higher professionalism corresponds to higher ratings. Authority refers to the influence of the media; for example, in car reviews, professional car forums have higher authority than social media.

[0042] Based on the above evaluation indicators and their corresponding weights, the target distribution for a single sample of public opinion information can be calculated. : ; in, They represent the first i The number of views, interaction rate, professionalism, authority, and timeliness of each sample of public opinion information.

[0043] Since view count is the easiest data to obtain and relatively accurate, and it has the greatest influence in product review public opinion, it can be used as the proposed distribution. Therefore, for a single public opinion information sample... The proposal distribution for: ; So, a single piece of public opinion information sample Importance weight for: .

[0044] Using the above method, the importance weight of each public opinion information sample in each group of evaluation data can be calculated. Then, each group of evaluation data is sorted from high to low according to the importance weight. A specified number of evaluation data from users with the highest importance weights are extracted from each group of evaluation data as a sampling data group. This results in 4 sampling data groups: text data group, image data group, chart data group, and video data group.

[0045] The target distribution of each public opinion information sample is calculated by multi-dimensional evaluation indicators, and the resulting importance weights are highly accurate, improving the accuracy of data sampling results and ensuring that the sampled data sets are accurate and effective.

[0046] In some embodiments, public opinion analysis is performed on each group of sampled data to obtain initial intermediate information, including: Send the text data set to the text analysis model to obtain information on sentiment intensity and evaluation focus; Send the image data set to the image analysis model to obtain data on product parts' attention and usage scenarios; Send the chart data set to the chart analysis model to obtain market sales attention and driving range attention; The video data sets are sent to the video analysis model to obtain key test steps, quantitative data on test results, and sentiment data.

[0047] In public opinion analysis, four independent analysis models can be used, all running in parallel and capable of communication, information sharing, and mutual support. The four models have clearly defined roles, collaboratively processing data in parallel to output initial intermediate information for public opinion analysis. Specifically: The text analysis model determines the sentiment (positive / neutral / negative) of each piece of text data, detects sarcasm / metaphors / internet slang, identifies the intensity of sentiment (1-10 scale), and captures customer evaluation focus points (exterior, interior, handling performance, vehicle quality, level of intelligence, etc.). Simultaneously, based on the data source labeling information, the text analysis results are shared with three other analysis models to assist them in conducting in-depth analysis of data from the same source.

[0048] Image analysis models are used to deeply analyze image data, helping to understand the emotional messages conveyed by customers. For example, by recognizing vehicle parts, it can analyze customer attention levels for areas such as the front, headlights, interior, and rear. For usage scenario data, such as vehicle testing scenario recognition, it can determine whether a customer is testing on a professional racetrack, highway, city road, or muddy road, thus analyzing the driving scenarios that interest them most. By obtaining information such as product part attention levels and usage scenario data, it is possible to better understand customer emotional messages. Simultaneously, based on data source labeling information, image analysis results are shared with three other analysis models to assist them in conducting in-depth analysis of data from the same source.

[0049] Chart analysis models are used for in-depth analysis of chart data, identifying chart types (bar charts, tables, line charts, pie charts, etc.) and extracting core data. For example, a bar chart of monthly sales figures for a vehicle model can determine customer attention to market sales and popularity. A comparison chart of driving range with competing models can determine customer attention to driving range. By analyzing charts to obtain information such as market sales attention and driving range attention, it is possible to help uncover future product trends and extract valuable information for consumers and automakers. Simultaneously, based on data source channel labeling information, the chart analysis results are shared with three other analysis models to assist them in conducting in-depth analysis of data from the same source.

[0050] Video analytics models are used for in-depth analysis of video data. Videos typically contain image frames, audio, and subtitles, making them the most information-rich and intuitive data type. The video analytics agent identifies key testing scenarios (e.g., braking tests, acceleration tests, cornering tests, road condition tests, comfort tests, smart cockpit tests, intelligent driving performance tests, etc.) from the video. It extracts key quantitative data (e.g., 0-100 km / h acceleration data, braking distance data, etc.) from the video. It also gathers customer sentiment from the video (subjective feelings such as "sportiness," "comfort," and "luxury" conveyed by customers in the video and audio, as well as potential information such as purchase intent). Simultaneously, based on data source labeling information, the video analytics results are shared with three other analytics models to assist them in conducting in-depth analysis of data from the same source.

[0051] By using corresponding analysis models to process different data types in parallel, the specialization of each analysis model can be fully utilized, avoiding the problem of being comprehensive but not precise due to using only a single analysis model. This results in a more comprehensive, in-depth, and credible public opinion analysis report.

[0052] Finally, the obtained initial intermediate information is integrated. If a user-defined outline or template is available, the initial intermediate information, including the summary, data, citations, and charts, will be organized according to the user-specified structure to generate a final public opinion report. If the user does not specify a report structure, a final public opinion report will be generated based on the initial intermediate information, including the summary, data, citations, and charts, using the default style.

[0053] Please refer to Figure 5 , Figure 5 The structural block diagram of a product public opinion report generation device provided in this application should be understood to be related to... Figure 1 The method embodiment executed in this document corresponds to the method described above, and is capable of performing the steps involved in the aforementioned method. The specific functions of this device can be found in the description above; to avoid repetition, detailed descriptions are appropriately omitted here. This device includes, but is not limited to: The data acquisition module 101 is used to collect and preprocess the evaluation data to obtain multiple sets of evaluation data based on data type classification; The sampling module 102 is used to sample each set of evaluation data based on the importance sampling principle to obtain a sample data set for each data type. The public opinion analysis module 103 is used to perform public opinion analysis on each group of sampled data to obtain initial intermediate information. The report generation module 104 is used to generate public opinion reports based on initial intermediate information.

[0054] In the technical solution of this application embodiment, based on the principle of importance sampling, effective data is extracted from massive resources, and then public opinion analysis is performed on the effective data, which can effectively improve processing efficiency, avoid overload of similar data, reduce the consumption of computing resources, and improve report generation efficiency and report quality.

[0055] According to some embodiments of this application, the data acquisition module 101 is specifically used to collect evaluation data based on user-input demand information; clean the collected evaluation data and mark its data type and source; classify the evaluation data based on its data type to obtain multiple sets of evaluation data, including text, images, charts and videos.

[0056] According to some embodiments of this application, the sampling module 102 is specifically used to calculate the importance weight of each public opinion information sample in each group of evaluation data based on the importance sampling principle; sort each group of evaluation data from high to low based on the importance weight; and obtain the top M public opinion information samples as the sampling data group according to the sorting result, where M is a set value.

[0057] According to some embodiments of this application, the specific calculation process of importance weight includes: calculating the target distribution of a single public opinion information sample based on pre-set evaluation indicators and corresponding weights. The proposal distribution is determined based on the most readily available and weighted evaluation metrics. The importance weight of a single public opinion information sample is determined based on the target distribution and the proposal distribution. : .

[0058] According to some embodiments of this application, the target distribution of a single public opinion information sample is represented as follows: ;in, Indicates the first i Sample of public opinion information Target distribution These are samples of a single piece of public opinion information. The corresponding evaluation metrics are calculated, including viewership, interaction rate, professionalism, authority, and timeliness. The weights corresponding to the evaluation metrics are respectively and .

[0059] According to some embodiments of this application, the public opinion analysis module 103 is specifically used to send text data sets to a text analysis model to obtain information on sentiment intensity and evaluation focus points; send image data sets to an image analysis model to obtain information such as product part attention and usage scenario data to assist in understanding customer sentiment information; send chart data sets to a chart analysis model to obtain information such as market sales attention and driving range attention to assist in exploring future product trends; and send video data sets to a video analysis model to obtain key testing links, quantitative data of test results, and sentiment trend data.

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below. In some embodiments, please refer to... Figure 6 , Figure 6 A structural diagram of the product public opinion report generation system. This system includes: The aforementioned product public opinion report generation device and coordination and command module 105, wherein: The coordination and command module 105 is used to decompose the public opinion report generation task into the following sub-tasks based on the user's input requirements: data collection and preprocessing, data sampling, public opinion analysis, and report generation. These sub-tasks are then sequentially distributed to the respective processing modules in the public opinion report generation device of the product for processing, and task scheduling is performed during the processing.

[0061] This system breaks down the complex task of generating product evaluation and public opinion analysis reports into a series of sub-tasks, including: requirements analysis, data acquisition and preprocessing, data sampling, data analysis, and report generation. This task decomposition fully leverages the expertise of each module, avoiding the problem of a single module being comprehensive but lacking depth. Under the unified management of the coordination and command module 105, a multi-module parallel strategy is used to efficiently generate more comprehensive, in-depth, and reliable product evaluation and public opinion analysis reports. This solves the problems of low efficiency in data processing capabilities of single modules, insufficient comprehensiveness in problem analysis, poor system robustness, and inability to quickly adapt to various new requirements.

[0062] The data acquisition module 101 includes three steps: demand analysis, data collection, and data preprocessing. The data acquisition module 101 receives instructions from the coordination and command module 105, which include information such as the target vehicle model, time range, amount of information, data type, and source output from the demand analysis. Based on the instructions, the data acquisition module 101 retrieves raw data related to public opinion needs from different channels in parallel and stores it in the raw data cache.

[0063] The public opinion analysis module 103 includes a text analysis model, an image analysis model, a chart analysis model, and a video analysis model. These models operate in parallel and can communicate and share information to complement each other, providing mutual support in analysis. The four analysis models have clearly defined roles and collaboratively process data in parallel, outputting initial intermediate information for public opinion analysis. For example, the text analysis model receives instructions from the coordination and command module 105 to perform in-depth analysis of text data and shares the text analysis results with the other three analysis models based on the data source channel marking information to assist them in conducting in-depth analysis of data from the same source.

[0064] The report generation module 104 receives instructions from the coordination and command module 105 and integrates the intermediate information output from the four analysis models. If a user-defined outline or template is provided, the module will organize the intermediate information, such as the summary, data, citations, and charts, according to the user-specified structure to generate a final public opinion report. If the user does not specify a report structure, the report generation module 104 will organize the intermediate information, such as the summary, data, citations, and charts, according to the default style to generate a final public opinion report.

[0065] The specific implementation methods of the coordination and command module 105 for task allocation, resource scheduling, progress monitoring and result fusion, conflict resolution, and inter-module communication are as follows: To adapt to the diverse data sources and broad analytical dimensions required for products like automotive public opinion monitoring, this system adopts a "hybrid control architecture." This architecture involves a centralized decision-making process by the coordination and command module 105, while various processing modules execute tasks in a distributed manner. The coordination and command module 105 is responsible for overall decision-making, including task allocation, resource allocation, progress monitoring and result fusion, and conflict resolution between modules. Other processing modules are responsible for executing sub-tasks locally and interact with the coordination and command module 105 at key nodes through status reporting interfaces.

[0066] In the task initialization and decomposition process, the system receives user requests, and the coordination and command module 105 decomposes these requests into a chain of executable sub-tasks. Taking the user request "Generate a quarterly report on public opinion analysis of the GAC Aion RT model from July to September 2025" as an example, the core parameters of the request are as follows: Core parameters = {"brand":"GAC Aion","model":"RT", "time_range":("2025-07-01","2025-09-30"), "platforms":["Autohome", "Dongchedi", "Yiche", "Guazi Used Cars", "Douyin", "Xiaohongshu", "Weibo", "Bilibili"], "report_type":"quarterly", "analysis_dimensions":["Sentiment Analysis","Hot Topics","Risk Warning","User Profile","Purchase Intention","Competitor Comparison","Sales Statistics","Exterior Evaluation","Intelligent Assisted Driving Evaluation","Intelligent Cockpit Evaluation"] }

[0067] The coordination and command module 105 breaks down the overall public opinion analysis task into a chain of sub-tasks as follows: Sub-task chain = { {"task_id":"collect","type":"data collection","params":core parameters}, {"task_id":"preprocess","type":"data preprocessing","params":core parameters}, {"task_id":"sampling","type":"data sampling","params":core parameters}, {"task_id":"analyze","type":"data analysis","params":core parameters}, {"task_id":"generate","type":"report generation","params":core parameters} }

[0068] Task scheduling, or the coordination and command module 105, schedules each processing module (including data acquisition module 101, sampling module 102, public opinion analysis module 103, and report generation module 104) in the following sequence: "demand analysis - data collection - data preprocessing - data sampling - public opinion analysis - report generation". The next step is automatically started after the previous step is completed.

[0069] The resource allocation strategy adopts a Ray-based resource-aware load balancing scheduling strategy. The coordination and command module 105 automatically allocates subtasks to idle processing nodes based on subtask type, data type, CPU / GPU load status of each processing module, and task queue length. This enables parallel execution of tasks, enhances data processing capabilities, and improves overall system efficiency.

[0070] Progress monitoring and result integration: The coordination and command module monitors the task progress of each processing module in real time through the "status reporting interface". After each stage is completed, each processing module sends the task execution status to the result message queue. The coordination and command module reads the sub-task progress status from the result message queue and triggers the next stage based on the progress status. The coordination and command module summarizes the results of the public opinion analysis module 103 and sends them to the report generation module 104. The report generation module 104 integrates the results and generates structured content according to the requirements of the public opinion report outline.

[0071] Conflict resolution strategies: Because this system employs parallel processing, resource contention or task conflicts arise between different processing modules. Conflict resolution primarily utilizes a combination of the following strategies: Priority Mechanism: Priority levels are assigned to tasks (e.g., P0 for urgent tasks, P1 for normal tasks). The coordination and command module prioritizes scheduling high-priority tasks, while low-priority tasks can be preempted. Timeout Retry Mechanism: If any processing module times out or fails to execute a subtask, the coordination and command module reassigns the task to another idle module. Data Consistency Mechanism: Distributed locks are used to prevent multiple processing modules from simultaneously modifying shared data.

[0072] Communication Strategy: Communication is fundamental to the coordination between the command and control module and the various processing modules. The system employs a publish-subscribe asynchronous communication model to meet the requirements of low latency, high reliability, and scalability. The communication protocol used is the lightweight RedisPub / Sub scheme. Specifically: the command and control module acts as the "publisher," issuing task instructions to designated processing modules. Each processing module acts as a "subscriber," subscribing to topics of interest and receiving tasks.

[0073] The format of the task instructions issued by the coordination and command module to each processing module is as follows: { "message_type":"task_assignment", / / Message type (task assignment / status query / resource update, etc.) "task_id":"task-20250820-165220-0005", / / Unique identifier for the task "content":"Analyze this chart", / / Task content "params":{ / / Task parameters (structured configuration) "model_name":"Qwen3-235B", "gpu_id":0, "max_tokens":4096, "timeout":300 / / Timeout duration (seconds) }, "priority":"P0", / / Task priority "timestamp":1716182400 / / Timestamp (used for time synchronization)}.

[0074] The format of the status feedback content published by each processing module to the coordination and command module is as follows: { "message_type":"status_update", "agent_id":"worker-008", "task_id":"task-20250820-165220-0005", "status":"completed", / / status (idle / busy / completed / failed) "result":"......", / / Task result "metrics":{ / / Task execution performance metrics "execution_time":120, "gpu_utilization":85.2 }, "error":null / / Error message }

[0075] The entire process of automating the generation of public opinion reports is managed uniformly by the coordination and command module, while other processing modules focus on a single sub-task. On the one hand, this avoids the problem of being all-encompassing but not specialized due to a single module handling all tasks; on the other hand, it improves the robustness of the entire system, and the use of parallel tasks enables efficient execution of the overall process, improving the utilization rate of computing resources.

[0076] This application provides an electronic device including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to cause the electronic device to perform the methods in any of the aforementioned optional implementations.

[0077] This application provides a readable storage medium storing computer program instructions, which, when read and executed by a processor, perform the methods in any of the aforementioned optional implementations.

[0078] This application provides a computer program product comprising a computer program that, when executed by a processor, performs the method in any of the aforementioned optional implementations.

[0079] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0081] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0082] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0084] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0085] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for generating a product public opinion report, characterized in that, The method includes: The evaluation data is collected and preprocessed to obtain multiple sets of evaluation data based on data type classification; Based on the importance sampling principle, data is sampled for each set of evaluation data to obtain sample data sets for each data type; Perform public opinion analysis on each group of sampled data to obtain initial intermediate information; A public opinion report is generated based on the initial intermediate information.

2. The product public opinion report generation method according to claim 1, characterized in that, The process of collecting and preprocessing evaluation data to obtain multiple sets of evaluation data based on data type classification includes: Evaluation data is collected based on user-inputted requirements. The collected evaluation data is cleaned and its data type and source are marked. The evaluation data is classified based on the data type to obtain multiple sets of evaluation data. The data type includes text, images, charts, and videos.

3. The product public opinion report generation method according to claim 1, characterized in that, The method of sampling data for each set of evaluation data based on the importance sampling principle to obtain sample data sets for each data type includes: The importance weight of each public opinion information sample in each set of evaluation data is calculated based on the aforementioned importance sampling principle. Based on the aforementioned importance weights, each set of evaluation data is sorted in descending order; The top M public opinion information samples are obtained based on the sorting results as the sampling data group, where M is a set value.

4. The product public opinion report generation method according to claim 3, characterized in that, The calculation of the importance weight of each public opinion information sample in each set of evaluation data based on the importance sampling principle includes: The target distribution of a single public opinion information sample is calculated based on pre-defined evaluation indicators and corresponding weights. ; The proposal distribution was determined based on the most readily available and weighted evaluation metrics. ; The importance weight of each individual public opinion information sample is determined based on the target distribution and the proposal distribution. : 。 5. The product public opinion report generation method according to claim 4, characterized in that, The target distribution of a single public opinion information sample is represented as follows: ； in, Indicates the first i Sample of public opinion information Target distribution These are samples of a single piece of public opinion information. The calculation results of the corresponding evaluation indicators, including viewership, interaction rate, professionalism, authority, and timeliness. The weights corresponding to the evaluation indicators are respectively and .

6. The product public opinion report generation method according to claim 1, characterized in that, The sampled data sets include text data sets, image data sets, chart data sets, and video data sets. The process of performing public opinion analysis on each sampled data set to obtain initial intermediate information includes: The text data set is sent to a text analysis model to obtain information on sentiment intensity and evaluation focus; The image data set is sent to the image analysis model to obtain data on product part attention and usage scenario; The chart data set is sent to the chart analysis model to obtain market sales attention and driving range attention; The video data set is sent to the video analysis model to obtain key test steps, quantitative data of test results, and sentiment data.

7. A product public opinion report generation device, characterized in that, The device includes: The data acquisition module is used to collect and preprocess the evaluation data to obtain multiple sets of evaluation data based on data type classification; The sampling module is used to sample data for each set of evaluation data based on the importance sampling principle, and obtain sample data sets for each data type. The public opinion analysis module is used to perform public opinion analysis on each group of sampled data to obtain initial intermediate information. The report generation module is used to generate a public opinion report based on the initial intermediate information.

8. A product public opinion report generation system, characterized in that, The system includes: the product public opinion report generation device and the coordination and command module as described in claim 7, wherein: The coordination and command module is used to decompose the public opinion report generation task into the following sub-tasks based on the user's input requirements: data collection and preprocessing, data sampling, public opinion analysis, and report generation. These sub-tasks are then sequentially distributed to the respective processing modules in the product's public opinion report generation device for processing, and task scheduling is performed during the processing.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the product public opinion report generation method according to any one of claims 1 to 6.

10. A readable storage medium, characterized in that, The readable storage medium stores computer program instructions, which are read and executed by a processor to perform the product public opinion report generation method according to any one of claims 1 to 6.

11. A computer program product, characterized in that, The computer program product includes a computer program, which, when executed by a processor, performs the product public opinion report generation method according to any one of claims 1-6.

Images