Public opinion analysis method and device, electronic equipment and storage medium

Abstract

The invention relates to the field of public opinion processing, in particular to a public opinion analysis method and device, electronic equipment and a storage medium, according to the method provided by the embodiment of the invention, a closed loop is formed in the whole process of obtaining a public opinion analysis result from public opinion data, and the continuity and accuracy of public opinion analysis are ensured. And in the analysis process, performing layered identification on the target public opinion data to obtain a public opinion data set corresponding to each product model. Details and characteristics of public opinion data can be deeply mined, and richer information is provided for subsequent public opinion analysis. And then public opinion analysis is carried out based on the public opinion data set corresponding to each product model, so that the public opinion conditions of the product in different aspects can be more comprehensively known. And through comprehensive analysis of data of different dimensions, potential problems and risks can be found.

Description

Technical Field

[0001] The present invention relates to the field of public opinion processing, and in particular to a public opinion analysis method, device, electronic equipment and storage medium. Background Art

[0002] The current general model performs poorly in public opinion analysis in specific fields, lacks deep customization capabilities, and is difficult to accurately judge emotions and functional evaluations. The original public opinion data comes from various sources and formats, and there is a problem of label abuse, which makes data classification and problem attribution difficult. Public opinion data is usually only used as auxiliary decision-making information. The existing analysis scheme lacks deep mining and closed-loop processing, resulting in the inability to discover potential problems and risks. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a public opinion analysis method, device, electronic device and storage medium to solve the problem that existing public opinion analysis solutions lack deep mining and closed-loop processing, resulting in the inability to discover potential problems and risks.

[0004] In a first aspect, an embodiment of the present invention provides a method for analyzing public opinion, the method comprising:

[0005] Obtain the original public opinion data corresponding to each product model included in the target product category;

[0006] Analyze the original public opinion data to obtain target public opinion data carried in the original public opinion data;

[0007] Performing hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each of the product models, wherein the public opinion data set includes data of the product model in different dimensions;

[0008] Public opinion analysis is performed based on the public opinion data sets corresponding to each of the product models to obtain a public opinion analysis result.

[0009] Furthermore, the analyzing the original public opinion data to obtain the target public opinion data carried in the original public opinion data includes:

[0010] Identify the data type corresponding to the original public opinion data, wherein the data type includes but is not limited to image type, video type and text type;

[0011] The original public opinion data is processed using a preset processing strategy corresponding to the data type to obtain the target public opinion data.

[0012] Furthermore, the target public opinion data is hierarchically identified to obtain public opinion data sets corresponding to each product model, including:

[0013] Extracting product public opinion data corresponding to the product model from the target public opinion data;

[0014] The product public opinion data is hierarchically identified according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model.

[0015] Furthermore, the product public opinion data is hierarchically identified according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model, including:

[0016] Acquire a first preset dimension based on the preset hierarchical identification strategy;

[0017] Extracting first data matching the first preset dimension from the product public opinion data;

[0018] Determine a second preset dimension at a level below the first preset dimension based on the preset hierarchical identification strategy;

[0019] Extracting second data matching the second preset dimension from the first data;

[0020] Based on the preset hierarchical identification strategy, determining whether there is a third preset dimension located at a level below the second preset dimension;

[0021] If the third preset dimension exists, third data matching the third preset dimension is extracted from the second data until data corresponding to the preset dimension of each level in the preset hierarchical identification strategy is obtained, and the public opinion data set is constructed based on the data corresponding to the preset dimension of each level.

[0022] Furthermore, the public opinion analysis is performed based on the public opinion data set corresponding to each product model to obtain the public opinion analysis result, including:

[0023] Performing public opinion analysis based on the public opinion data set corresponding to each of the product models to obtain hot issues and warning issues corresponding to the product models;

[0024] Clustering the hot issues according to preset indicators to obtain problem clusters, and determining the public opinion data set with the warning issues as the target public opinion data set;

[0025] Analyze the target public opinion data set to obtain the cause of the warning problem;

[0026] The problem cluster and the cause of the warning problem are used as the public opinion analysis result.

[0027] Furthermore, the analysis based on the target public opinion data set to obtain the cause of the warning problem includes:

[0028] Identify the public opinion data set to obtain a product identification code and a functional description corresponding to the warning problem;

[0029] Obtaining an event list corresponding to the product identification code from the cloud, wherein the event list includes multiple events and product-related data when the events occur;

[0030] Acquire target data related to the functional description from the product-related data;

[0031] The target data is analyzed to determine the cause of the warning problem.

[0032] Furthermore, after clustering the hot issues according to preset indicators to obtain issue clusters, the method further includes:

[0033] Determine the function to be optimized of the product corresponding to the product model based on the problem cluster, and obtain the optimization solution corresponding to the function to be optimized;

[0034] Testing the effectiveness of the optimization scheme;

[0035] If the optimization scheme is effective, the optimization scheme is used to optimize the functions of the product corresponding to the product model.

[0036] In a second aspect, an embodiment of the present invention provides a public opinion analysis device, the device comprising:

[0037] An acquisition module is used to acquire the original public opinion data corresponding to each product model included in the target product category;

[0038] An analysis module, used to analyze the original public opinion data to obtain target public opinion data carried in the original public opinion data;

[0039] An extraction module, used to perform hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each product model, wherein the public opinion data set includes data of the product model in different dimensions;

[0040] The analysis module is used to perform public opinion analysis based on the public opinion data set corresponding to each product model to obtain a public opinion analysis result.

[0041] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0042] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method of the first aspect or any corresponding embodiment thereof.

[0043] The method provided in the embodiment of the present application has the following beneficial effects:

[0044] The method provided in the embodiment of the present application forms a closed loop from the entire process of obtaining public opinion data to the public opinion analysis results, ensuring the consistency and accuracy of the public opinion analysis. During the analysis process, the target public opinion data is layered and identified to obtain the public opinion data sets corresponding to each product model. It is possible to deeply explore the details and characteristics of the public opinion data and provide richer information for subsequent public opinion analysis. Then, based on the public opinion data sets corresponding to each product model, public opinion analysis can be performed to more comprehensively understand the public opinion situation of the product in different aspects. And through the comprehensive analysis of data of different dimensions, potential problems and risks can be discovered. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 is a flowchart of a public opinion analysis method according to some embodiments of the present invention;

[0047] Figure 2 is a flowchart of a public opinion analysis method according to some embodiments of the present invention;

[0048] Figure 3 is a structural block diagram of a public opinion analysis device according to an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of the hardware structure of the electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0051] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below.

[0052] Herein, unless explicitly stated, executing a step “in response to A” does not mean executing the step immediately after “A” but may include one or more intermediate steps.

[0053] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition, use, storage or deletion of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0054] It is understandable that before using the technical solutions disclosed in the various embodiments of the present disclosure, the types, scopes of use, usage scenarios, etc. of the information involved in the present disclosure should be informed to relevant users and their authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations. The relevant users may include any type of right holders, such as individuals, enterprises, and groups.

[0055] For example, in response to receiving an active request from a user, a prompt message is sent to the relevant user to clearly prompt the relevant user that the operation requested to be performed will require obtaining and using the information of the relevant user, so that the relevant user can independently choose whether to provide information to software or hardware such as an electronic device, application, server or storage medium that executes the operation of the technical solution of the present disclosure based on the prompt message.

[0056] As an optional but non-limiting implementation, in response to receiving an active request from a relevant user, a prompt message is sent to the relevant user, for example, in the form of a pop-up window, in which the prompt message may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to select "agree" or "disagree" to provide information to the electronic device.

[0057] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0058] According to an embodiment of the present invention, a public opinion analysis method, apparatus, electronic device and storage medium are provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0059] In this embodiment, a public opinion analysis method is provided. Figure 1 is a flow chart of a public opinion analysis method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0060] Step S101, obtaining original public opinion data corresponding to each product model included in the target product category.

[0061] In an embodiment of the present application, first, the target product category for which public opinion data is to be obtained is clearly defined, such as compact cars, mid-size SUVs, etc. in the automobile classification. Secondly, the specific product models under this category are determined, such as a specific model of a certain brand. Then the data source is determined, which can be a news website, television media, automobile forum, social media platform, e-commerce platform, etc. Finally, using a web crawler tool, relevant news and comment data can be automatically captured from a specified website. By setting specific keywords, page ranges, and crawling frequencies, the latest public opinion data can be ensured. Alternatively, using professional social media monitoring tools, discussions of specific keywords, topics, or brands on social media platforms can be monitored in real time. In this way, news, blogs, comments, and other data collected from various data sources are used as original public opinion data.

[0062] Step S102, analyzing the original public opinion data to obtain the target public opinion data carried in the original public opinion data.

[0063] In an embodiment of the present application, analyzing the original public opinion data to obtain the target public opinion data carried in the original public opinion data includes the following steps: identifying the data type corresponding to the original public opinion data, where the data type includes but is not limited to image type, video type and text type. Processing the original public opinion data using the preset processing strategy corresponding to the data type to obtain the target public opinion data.

[0064] Specifically, for image types, the corresponding processing strategy may be: using image recognition technology to extract key elements, logos or text information in the image. For example, for a picture of a car, the model, brand logo, etc. may be identified. For an image containing text, optical character recognition (OCR) technology may be used to extract the text for analysis.

[0065] For video types, the corresponding processing strategy can be: use video analysis tools to extract key frames, video titles, description information, etc. For the audio part of the video, voice recognition can be performed to convert the voice content into text for analysis. At the same time, metadata information such as video tags and classifications can be analyzed to understand the theme and content of the video.

[0066] For text types, the corresponding processing strategy can be: use natural language processing (NLP) technology to perform operations such as word segmentation, part-of-speech tagging, and named entity recognition on the text to better understand the content and structure of the text. It can remove irrelevant information such as stop words and punctuation marks, and extract key sentences and paragraphs.

[0067] Target public opinion data includes: topics, key points, author intentions, emotions, conclusions or suggestions, background information, etc. For example: by analyzing the title, opening paragraph, and key sentences of the text, extract the theme or central idea of ​​the post discussion. For example, if the post is about the quality problem of a certain car model, the theme can be determined as "Discussion of car model quality problems." Or extract important information, data, facts, or arguments from the text. You can focus on numbers, statistics, expert opinions, user experience, and other content. For example, if a post mentions that a certain car model has a long braking distance, this can be used as a key point. Alternatively, analyze the author's wording, tone, and expression to infer the message the author wants to convey through the post. For example, if the author strongly criticizes the after-sales service of a certain car model, his intention may be to call on the manufacturer to improve its after-sales service.

[0068] Step S103, performing hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each product model, wherein the public opinion data set includes data of the product model in different dimensions.

[0069] In the embodiment of the present application, the target public opinion data is hierarchically identified to obtain the public opinion data set corresponding to each product model, including the following steps A1-A2:

[0070] Step A1, extracting product public opinion data corresponding to the product model from the target public opinion data.

[0071] Specifically, first, determine the specific characteristics of each product model in the target product category. These characteristics include brand name, model name, specific configuration code or version number, etc. For example, for automobile products, the product model may be "Brand A Model B 2.0T Luxury Edition", where the brand, model and configuration information are all important identification features. A keyword library pre-established based on the characteristics of the product model can be determined. The keyword library can include the full name, abbreviation, abbreviation of the product model and common words related to the model. For example, for the above-mentioned automobile product model, keywords may include "Brand A", "Model B", "2.0T", "Luxury Edition", "Brand A Model B", etc.

[0072] For text data in the target public opinion data, natural language processing technology is used for text analysis. Scan the text content line by line to find whether it contains words in the keyword library. If matching keywords are found, the text is marked as public opinion data related to a specific product model. For example, if a news report mentions "how does brand A model B perform in the market", through keyword matching, it can be determined that this report is related to the product model and extracted as product public opinion data. For image and video data, image recognition and video analysis technology can be used. For example, the car brand logo and model features in the picture are identified through image recognition technology. If they match the target product model, the image is marked as relevant public opinion data. For videos, the title, description, keyframe and other information of the video can be analyzed to find whether it contains keywords for the product model. For example, if the title of a video is "Brand A Model B Test Drive Experience", through keyword matching, it can be determined that this video is related to the product model and extracted as product public opinion data.

[0073] Step A2: hierarchically identify the product public opinion data according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model.

[0074] Specifically, the product public opinion data is hierarchically identified according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model, including: obtaining a first preset dimension based on the preset hierarchical identification strategy, and extracting first data matching the first preset dimension from the product public opinion data; determining a second preset dimension located at a level below the first preset dimension based on the preset hierarchical identification strategy; extracting second data matching the second preset dimension from the first data; determining whether there is a third preset dimension located at a level below the second preset dimension based on the preset hierarchical identification strategy; if there is a third preset dimension, extracting third data matching the third preset dimension from the second data, until data corresponding to the preset dimensions at each level in the preset hierarchical identification strategy are obtained, and a public opinion data set is constructed based on the data corresponding to the preset dimensions at each level.

[0075] First, determine the first preset dimension from the preset hierarchical identification strategy. The first preset dimension can be a field, such as smart cockpit and smart driving. From the overall product public opinion data, filter out public opinion content related to "smart cockpit" or "smart driving" to form the first data. For example, it contains content that evaluates the comfort and convenience of the vehicle's smart cockpit, as well as content that evaluates the safety and reliability of smart driving assistance functions.

[0076] Secondly, according to the preset hierarchical identification strategy, the second preset dimension at the next level of the first preset dimension is determined, and the second preset dimension can be the vehicle function. The vehicle function can be further subdivided into the multimedia system, seat adjustment function, etc. in the smart cockpit, and the automatic parking function and adaptive cruise function in smart driving. From the first data, extract the public opinion data related to the specific vehicle function to form the second data. For example, from the first data related to the smart cockpit, extract the public opinion content that specifically discusses the smoothness of multimedia system operation, sound quality effects, etc.

[0077] Then, according to the preset hierarchical identification strategy, it is determined whether there is a third preset dimension located at a level below the second preset dimension. The third preset dimension can be user behavior, such as complaints, suggestions, consultations, etc. If the third preset dimension exists, the third data matching the complaints, suggestions, and consultations are extracted from the second data. For example, from the second data related to the multimedia system function, the user's complaints about the function (such as poor sound quality, complicated operation, etc.), suggestions (such as adding functions, optimizing the interface, etc.) and consultations (such as how to use a specific function, etc.) are screened out.

[0078] In addition, further sentiment analysis can be performed on the third data corresponding to complaints, suggestions, and consultations. Specifically, for complaint data: each complaint can be analyzed one by one to extract key problem points. For example, for complaints about "poor sound quality", it can be summarized as "multimedia system sound quality problems"; for complaints about "complex operation", it can be summarized as "multimedia system operation complexity is high". Similar complaint points are merged, and duplicate content is removed to form a concise and clear list of complaint points. For suggestion data: the core content of each suggestion can be analyzed, for example, "adding functions" can be summarized as "multimedia system function expansion requirements", and "optimizing the interface" can be summarized as "multimedia system interface optimization suggestions". Similarly, similar suggestion points are merged to form a list of suggestion points. For consultation data: the main issues that can be consulted, such as "how to use a specific function" can be summarized as "consultation on the use of a specific function". Organize and form a list of consultation points.

[0079] Then, for complaint data, suggestion data, and consultation data, user sentiment can be judged in the following way: traverse the words of each public opinion text and judge whether it is in the positive sentiment dictionary or the negative sentiment dictionary. If the number of positive words in the text is greater than the number of negative words, it can be preliminarily judged as positive sentiment; otherwise, it is negative sentiment. Alternatively, different weights can be given to each positive word and negative word, and the comprehensive sentiment score is calculated based on the weights. If the score is greater than a certain threshold, it is judged as positive sentiment, and if it is less than a certain threshold, it is judged as negative sentiment.

[0080] Repeat the process and build the public opinion data set. According to the preset hierarchical identification strategy, continue to determine whether there is a preset dimension at the next level, and repeat the above data extraction process until all preset dimension levels are processed.

[0081] Finally, based on the data corresponding to the preset dimensions at each level, a public opinion dataset is constructed. This public opinion dataset has a clear hierarchical structure, from macro-domain classification to specific user emotions, and comprehensively reflects the public opinion status of products in different aspects. For example, in the field of smart cockpits, the public opinion dataset corresponding to a specific in-vehicle function may include users' specific evaluation content of the function under different emotional tendencies.

[0082] It should be noted that the preset hierarchical identification strategy is designed to classify and organize product public opinion data through gradually refined dimensions to form a hierarchical public opinion data set. This method helps to more deeply understand the performance of products in different aspects and user feedback, and provide a basis for targeted analysis and decision-making.

[0083] As an example, first, the collected public opinion data is input into the trained LLM model. The model analyzes the language patterns, keywords, publisher behavior and other features of the text to identify advertising content, such as obvious sales statements, excessive promotion of specific brands, etc.; identify sales content, such as vehicle promotion information, price discount advertisements, etc.; and identify spam content, such as meaningless character repetitions, small talk unrelated to automobile products, etc. After filtering, what remains is the effective public opinion data that truly reflects users' experience and opinions on automobile products.

[0084] Secondly, the valid public opinion data is input into the LLM model again, and a model library containing all the model information of the company is provided to the LLM model. The model cleans the public opinion data based on the model feature keywords in the text, the vehicle appearance described, performance parameters and other information, and accurately identifies and distinguishes the public opinion data of different models.

[0085] Furthermore, the public opinion data of a single model is input into the LLM model. The LLM model divides the public opinion data of the model according to the pre-set keywords and feature descriptions of different fields such as smart driving and smart cockpit. For example, if the comment mentioned that "the automatic parking function is not very accurate", the model will classify it into the field of smart driving; and "the in-car display screen is a bit slow to respond" will be classified into the field of smart cockpit.

[0086] Then, the valid information processed in the previous step is fed to the LLM model. The LLM model extracts relevant content from the text based on preset tags, such as suggestions, complaints, and consultations. For example, "I hope the manufacturer can improve the sensitivity of the brake system" will be extracted as a suggestion; "The engine of this car always breaks down" will be extracted as a complaint; "I would like to ask how much the maintenance cost of this car is" will be extracted as a consultation.

[0087] Finally, the classified content is input into the LLM model again. The LLM model summarizes the public opinion content in each field and extracts the key points. For example, for the public opinion content in the field of smart cockpits, it is summarized that "users report that the in-car display screen is slow to respond and the sound effect is poor." At the same time, the model will also judge the positive and negative emotions of each public opinion. For example, "The interior of this car is very luxurious and the seats are very comfortable" will be judged as a positive emotion; "The fuel consumption of the car is too high and I am not very satisfied" will be judged as a negative emotion.

[0088] In this way, the data obtained from the above steps are summarized to construct a public opinion dataset.

[0089] When conducting public opinion analysis, the language model (LLM) is used to analyze the cleaned public opinion data to accurately identify the public opinion data of different models. In this way, the information of the required model can be directly obtained to avoid the mixing of data of different models and inaccurate analysis. After determining the model, domain content clustering is performed. This step can split and classify the content corresponding to different fields in a single comment or video, such as distinguishing the content of different fields such as smart driving and smart cockpit. The advantage of this is that specific information in different fields can be accurately obtained, providing a basis for subsequent targeted analysis. Then LLM analysis is performed. Specifically, the content is first classified and extracted through LLM, such as extracting different types of content such as suggestions, complaints, and consultations. After that, the extracted content is further analyzed and judged, such as summarizing the key points of public opinion and judging positive and negative emotions. This process is analyzed by LLM at least twice to ensure the accuracy and comprehensiveness of the analysis.

[0090] In contrast, the existing solution of directly feeding the entire data to the large language model and letting the large language model directly draw conclusions has obvious shortcomings. In this solution, the large language model is prone to misunderstanding or bias in understanding the data, resulting in incorrect analysis results. This solution effectively avoids these problems through gradual classification and multiple analyses, and can provide more reliable results for public opinion analysis.

[0091] Step S104, performing public opinion analysis based on the public opinion data sets corresponding to each product model to obtain a public opinion analysis result.

[0092] In the embodiment of the present application, public opinion analysis is performed based on the public opinion data sets corresponding to each product model to obtain the public opinion analysis results, including the following steps B1-B4:

[0093] Step B1, performing public opinion analysis based on the public opinion data set corresponding to each product model to obtain hot issues and warning issues corresponding to the product model.

[0094] Specifically, first, collect the public opinion data sets corresponding to each product model to ensure the integrity and accuracy of the data. Extract keywords from the public opinion data set. These keywords can reflect the main issues that users are concerned about when discussing the product model. For example, in the public opinion data in the field of smart cockpits, keywords such as "screen resolution" and "speech recognition accuracy" may be extracted. Perform frequency statistics on the extracted keywords to determine which keywords appear the most times. Keywords with high frequency of occurrence often represent hot issues for the product model. For example, if the keyword "low screen resolution" appears frequently in the public opinion data, then the screen resolution problem is likely to be one of the hot issues for the product model. In addition, in addition to keyword frequency statistics, other factors can also be combined for comprehensive analysis. For example, considering the user's emotional tendency, if a certain issue has more negative comments, then this issue is also a hot issue.

[0095] As for early warning issues, they can be determined in the following ways: First, according to the domain and RAG document segmentation content, clarify which situations or factors are considered as early warning indicators of S / A level issues. These indicators may include the severity of the problem, the scope of impact, and the development trend. For example, problems involving safety hazards, which may lead to large-scale recalls, and problems that seriously damage the brand image. Secondly, collect the public opinion data sets corresponding to each product model to ensure the integrity and accuracy of the data. Perform preliminary processing on the data to remove irrelevant or noise information for subsequent analysis. Then extract keywords from the public opinion data set, and classify and filter the extracted issues according to the early warning indicators. Separate out matters that may constitute early warning issues. You can also continue to monitor and track the public opinion data set in real time and pay attention to the development and changes of the problem. Finally, combine the severity of the problem, the development trend and the relevant early warning indicators to conduct a comprehensive assessment and judgment to determine which problems have indeed reached the early warning level. When a problem is determined to be an early warning problem, issue an early warning signal in a timely manner to notify relevant departments and personnel to take corresponding measures. The early warning mechanism can include email notifications, SMS reminders, system pop-ups, etc.

[0096] Step B2: clustering hot issues according to preset indicators to obtain problem clusters, and determining the public opinion data set with warning issues as the target public opinion data set.

[0097] Specifically, according to actual needs and business priorities, corresponding weights are assigned to the three preset indicators of vehicle type, time, and field. For example, if the current focus is more on the differences between different vehicle types, the weight of the vehicle type can be relatively high; if the time factor has a greater impact on the problem, the weight of time can be appropriately increased.

[0098] First, based on the three preset indicators of vehicle model, time and field, construct the feature vector of each hot issue. For example, the feature vector of a hot issue can be expressed as [vehicle model code, timestamp, field code]. The feature vector can also be expanded according to the actual situation, such as adding keywords, sentiment tendencies and other relevant information of the problem. Then, according to the selected clustering algorithm, set the corresponding parameters. For example, for K-Means clustering, it is necessary to determine the number of clusters K; for DBSCAN, it is necessary to set parameters such as the neighborhood radius and the minimum number of samples. Initialize the clustering algorithm, such as randomly selecting the initial cluster center (for K-Means clustering) or determining the initial neighborhood range (for DBSCAN). Finally, take the feature vector of the hot issue as input and run the selected clustering algorithm. The algorithm will divide the hot issues into different clusters according to the preset indicators and parameters.

[0099] Step B3: Analyze the target public opinion data set to obtain the cause of the warning problem.

[0100] In the embodiment of the present application, the analysis is performed based on the target public opinion data set to obtain the cause of the warning problem, including the following steps:

[0101] Step B301, identify the public opinion data set to obtain the product identification code and the functional description corresponding to the warning problem.

[0102] Specifically, determine the rules for identifying product identification codes based on the characteristics of the product and common identification methods. For example, if the product is a car, the product identification code can be determined by the model name, VIN code (vehicle identification number), brand and model combination, etc. Carefully analyze each public opinion data to find information that may contain product identification codes. Verify and confirm the possible identification codes found based on the established identification rules.

[0103] For automobile products, early warning issues may include safety performance issues, quality defects, poor after-sales service, etc. Therefore, we analyze the content related to early warning issues in the public opinion data set. We find users' specific descriptions and opinions on the issues, as well as the product functions involved. We extract key functional description keywords from the public opinion content related to the early warning issues. These keywords can reflect the product functions or features involved in the problem. Based on the extracted keywords, we integrate the functional descriptions corresponding to the early warning issues.

[0104] Step B302, obtaining an event list corresponding to the product identification code from the cloud, wherein the event list includes multiple events and product-related data when the events occur.

[0105] Specifically, construct a request to obtain a list of events corresponding to the product identification code. Use the product identification code as one of the request parameters. For example, in the automotive field, if the VIN code is known, you can construct an HTTP request to pass the VIN code to the cloud server. Check the response status code of the request to ensure that the request is successful. If the response is successful, parse the returned event list data. The event list contains multiple events, each of which may have information such as the time of occurrence, event description, and product-related data.

[0106] Step B303, obtaining target data related to the function description from the product related data.

[0107] Specifically, according to the data type determined in the functional description, the target data related to the data is filtered out from the product-related data according to the data type. For example, if the engine speed data needs to be obtained, the data of the engine speed sensor can be filtered out from the sensor data obtained from the OBD reading device or the vehicle application. For vehicle driving video data, the video clips related to the time when the warning problem occurred can be filtered out according to the timestamp or event marker. For example, if the warning problem occurs within a specific time period, the video clips within that time period can be filtered out from the video data of the driving recorder.

[0108] Then, the filtered target data is processed to better analyze and understand the problem. For example, the sensor data can be converted into units, cleaned, and processed for outliers to ensure the accuracy and reliability of the data. For vehicle driving video data, video editing, image enhancement, key frame extraction, and other processing can be performed to more intuitively observe the situation when the problem occurs. For the engine operation data, data analysis and visualization processing can be performed, such as drawing speed-torque curves, fuel consumption-time curves, etc., to better understand the engine's operating status. Finally, the filtered and processed target data is stored in a suitable storage medium for subsequent analysis and processing.

[0109] Step B304, analyzing the target data to determine the cause of the warning problem.

[0110] Specifically, key features are extracted from the target data. For example, for engine speed data, features such as the mean, standard deviation, and rate of change of the speed can be extracted; for vehicle driving video data, features such as the vehicle's position, speed, acceleration, and steering angle can be extracted. Use the extracted key features to train a prediction model, such as a regression model, a classification model, or a clustering model. Model training can be performed using historical data to learn data patterns and features under normal circumstances. Use the trained model to perform early warning detection on new data. When a warning signal is detected, the warning mechanism is triggered.

[0111] For situations where an early warning is triggered, further data analysis is performed to determine the cause of the early warning problem. This can be achieved in the following ways: ① Compare the early warning data with the normal data to find out the differences and anomalies. ② Use professional fault diagnosis tools and techniques to check and diagnose the engine or other related systems to determine if there is a fault. ③ Carefully observe and analyze the vehicle driving video data to find out whether there are abnormal situations such as vehicle collisions and driver violations. ④ Correlate and analyze the early warning data with other related data, such as sensor data, vehicle status data, etc., to find possible causes.

[0112] Step B4, taking the problem cluster and the causes of the warning problem as the public opinion analysis results.

[0113] The method provided in the embodiment of the present application clusters hot issues according to preset indicators, and a large number of hot issues can be classified and sorted to reduce the complexity and confusion of the data. This makes the subsequent analysis more targeted. At the same time, by clustering to obtain problem clusters, similar problems can be classified into one category to clarify the type and characteristics of the problem. This helps technicians to locate the core of the problem more accurately, reduce the scope of problem analysis, and improve the efficiency of problem solving. In addition, the public opinion data set with early warning problems is determined as the target public opinion data set, so that key issues can be analyzed with concentration. This targeted analysis can dig deeper into the causes of early warning problems.

[0114] The method provided in the embodiment of the present application forms a closed loop from the entire process of obtaining public opinion data to public opinion analysis results, ensuring the consistency and accuracy of public opinion analysis. During the analysis process, the target public opinion data is layered and identified to obtain the public opinion data set corresponding to each product model. It is possible to deeply explore the details and characteristics of the public opinion data and provide richer information for subsequent public opinion analysis. Then, public opinion analysis is performed based on the public opinion data set corresponding to each product model, which can provide a more comprehensive understanding of the public opinion situation of the product in different aspects. And through the comprehensive analysis of data of different dimensions, potential problems and risks can be discovered.

[0115] In the embodiment of the present application, after clustering the hot issues according to the preset indicators to obtain the issue clusters, such as Figure 2 As shown, the method also includes:

[0116] Step S201, determining the function to be optimized of the product corresponding to the product model based on the problem cluster, and obtaining the optimization solution corresponding to the function to be optimized.

[0117] In an embodiment of the present application, problem clusters corresponding to product models are collected, and problem clusters are obtained by clustering hot issues. For example, for an electric car, the problem clusters may include battery life issues, car system freeze issues, etc. According to the frequency and severity of the problems in the problem clusters, the functions of the product to be optimized are determined. For example, if the battery life problem occurs frequently in the problem cluster and users complain a lot about it, then the battery life function can be determined as a function to be optimized.

[0118] Step S202, detecting the effectiveness of the optimization scheme.

[0119] In an embodiment of the present application, corresponding detection indicators are formulated according to the functions to be optimized and the optimization scheme to evaluate the effectiveness of the optimization scheme. For example, for the optimization scheme for battery life problems, detection indicators such as battery life, charging speed, and power consumption can be formulated. Then, after establishing a simulated experimental environment, the optimization scheme is experimentally tested to collect relevant data and indicators. For example, for the optimization scheme for battery life problems, multiple experimental tests can be performed to record indicators such as battery life, charging speed, and power consumption under different usage scenarios. Analyze the data and indicators collected from the experimental tests to evaluate the effectiveness of the optimization scheme.

[0120] Step S203: If the optimization solution is effective, the optimization solution is used to optimize the function of the product corresponding to the product model.

[0121] In the embodiments of the present application, a detailed optimization plan is generated according to the specific content of the optimization plan, and the optimization goals, steps and schedule are clearly defined. For example, for the optimization plan for battery life issues, specific optimization steps such as replacing large-capacity batteries, optimizing system power management, and reducing hardware power consumption can be formulated, and the completion time node and management object of each step can be determined. According to the optimization plan, the optimization plan is gradually implemented to optimize the functions of the products corresponding to the product model.

[0122] The method provided in the embodiment of the present application determines the functions to be optimized of the product through problem clusters, and can accurately find the specific problems that exist in the product under a specific model based on actual user feedback and market performance. Avoid blind product improvements and make the optimization work more targeted. After determining the functions to be optimized, obtain the corresponding optimization plan. For different functions to be optimized, you can formulate personalized optimization plans and perform effectiveness testing. The effectiveness testing can use simulation experiments, data analysis, user feedback collection and other means to verify the actual effect of the optimization plan from different angles, and provide a reliable basis for decision-making. If the optimization plan is effective and implemented, it can solve the user's pain points.

[0123] The method provided in the embodiment of the present application forms a closed-loop product optimization process, which can continuously promote product improvement and upgrading from problem discovery to solution formulation, effectiveness testing to implementation optimization.

[0124] In this embodiment, a public opinion analysis device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0125] This embodiment provides a public opinion analysis device, such as Figure 3As shown, including:

[0126] The acquisition module 301 is used to acquire the original public opinion data corresponding to each product model included in the target product category;

[0127] The analysis module 302 is used to analyze the original public opinion data to obtain the target public opinion data carried in the original public opinion data;

[0128] Extraction module 303, used to perform hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each product model, wherein the public opinion data set includes data of the product model in different dimensions;

[0129] The analysis module 304 is used to perform public opinion analysis based on the public opinion data sets corresponding to each product model to obtain a public opinion analysis result.

[0130] In an embodiment of the present application, the analysis module 302 is used to identify the data type corresponding to the original public opinion data, where the data type includes but is not limited to image type, video type and text type; the original public opinion data is processed using a preset processing strategy corresponding to the data type to obtain target public opinion data.

[0131] In an embodiment of the present application, the extraction module 303 is used to extract product public opinion data corresponding to the product model from the target public opinion data; the product public opinion data is hierarchically identified according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model.

[0132] In an embodiment of the present application, the extraction module 303 is used to obtain a first preset dimension based on a preset hierarchical identification strategy, extract first data matching the first preset dimension from the product public opinion data; determine a second preset dimension located at a level below the first preset dimension based on the preset hierarchical identification strategy; extract second data matching the second preset dimension from the first data; determine whether there is a third preset dimension located at a level below the second preset dimension based on the preset hierarchical identification strategy; if there is a third preset dimension, extract third data matching the third preset dimension from the second data, until data corresponding to the preset dimensions at each level in the preset hierarchical identification strategy are obtained, and a public opinion data set is constructed based on the data corresponding to the preset dimensions at each level.

[0133] In an embodiment of the present application, the analysis module 304 is used to perform public opinion analysis based on the public opinion data sets corresponding to each product model to obtain hot issues and warning issues corresponding to the product model; cluster the hot issues according to preset indicators to obtain problem clusters, and determine the public opinion data sets with warning issues as the target public opinion data sets; perform analysis based on the target public opinion data sets to obtain the causes of the warning issues; and use the problem clusters and the causes of the warning issues as the public opinion analysis results.

[0134] In an embodiment of the present application, the analysis module 304 is used to identify the public opinion data set to obtain a product identification code and a functional description corresponding to the warning problem; obtain an event list corresponding to the product identification code from the cloud, wherein the event list includes multiple events and product-related data when the events occur; obtain target data related to the functional description from the product-related data; and analyze the target data to determine the cause of the warning problem.

[0135] In an embodiment of the present application, the device also includes: an optimization module, which is used to determine the functions to be optimized of the products corresponding to the product model based on the problem cluster, and obtain the optimization plan corresponding to the functions to be optimized; detect the effectiveness of the optimization plan; if the optimization plan is effective, use the optimization plan to optimize the functions of the products corresponding to the product model.

[0136] See also Figure 4 , Figure 4 is a schematic diagram of the structure of an electronic device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system).

[0137] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0138] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0139] The memory 20 may 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 by the use of an electronic device based on the presentation of a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the electronic 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.

[0140] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0141] The electronic device further comprises a communication interface 30 for the electronic device to communicate with other devices or a communication network.

[0142] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0143] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for analyzing public opinion, characterized in that: The method comprises: Obtain the original public opinion data corresponding to each product model included in the target product category; Analyze the original public opinion data to obtain target public opinion data carried in the original public opinion data; Performing hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each of the product models, wherein the public opinion data set includes data of the product model in different dimensions; Public opinion analysis is performed based on the public opinion data sets corresponding to each of the product models to obtain a public opinion analysis result.

2. The method according to claim 1, characterized in that The analyzing the original public opinion data to obtain the target public opinion data carried in the original public opinion data includes: Identify the data type corresponding to the original public opinion data, wherein the data type includes but is not limited to image type, video type and text type; The original public opinion data is processed using a preset processing strategy corresponding to the data type to obtain the target public opinion data.

3. The method according to claim 1, characterized in that The step of performing hierarchical identification on the target public opinion data to obtain public opinion data sets corresponding to each product model includes: Extracting product public opinion data corresponding to the product model from the target public opinion data; The product public opinion data is hierarchically identified according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model.

4. The method according to claim 3, characterized in that: The step of performing hierarchical identification on the product public opinion data according to a preset hierarchical identification strategy to obtain a public opinion data set corresponding to the product model includes: Acquire a first preset dimension based on the preset hierarchical identification strategy; Extracting first data matching the first preset dimension from the product public opinion data; Determine a second preset dimension at a level below the first preset dimension based on the preset hierarchical identification strategy; Extracting second data matching the second preset dimension from the first data; Based on the preset hierarchical identification strategy, determining whether there is a third preset dimension located at a level below the second preset dimension; If the third preset dimension exists, third data matching the third preset dimension is extracted from the second data until the data corresponding to the preset dimension of each level in the preset hierarchical identification strategy is obtained, and the public opinion data set is constructed based on the data corresponding to the preset dimension of each level.

5. The method according to claim 1, characterized in that: The public opinion analysis is performed based on the public opinion data set corresponding to each product model to obtain the public opinion analysis result, including: Performing public opinion analysis based on the public opinion data set corresponding to each of the product models to obtain hot issues and warning issues corresponding to the product models; Clustering the hot issues according to preset indicators to obtain problem clusters, and determining the public opinion data set with the warning issues as the target public opinion data set; Analyze the target public opinion data set to obtain the cause of the warning problem; The problem cluster and the cause of the warning problem are used as the public opinion analysis result.

6. The method according to claim 5, characterized in that The analyzing based on the target public opinion data set to obtain the cause of the warning problem includes: Identify the public opinion data set to obtain a product identification code and a functional description corresponding to the warning problem; Obtaining an event list corresponding to the product identification code from the cloud, wherein the event list includes multiple events and product-related data when the events occur; Acquire target data related to the functional description from the product-related data; The target data is analyzed to determine the cause of the warning problem.

7. The method according to claim 5, characterized in that After clustering the hot issues according to preset indicators to obtain issue clusters, the method further includes: Determine the function to be optimized of the product corresponding to the product model based on the problem cluster, and obtain the optimization solution corresponding to the function to be optimized; Testing the effectiveness of the optimization scheme; If the optimization scheme is effective, the optimization scheme is used to optimize the functions of the product corresponding to the product model.

8. A public opinion analysis device, characterized in that: The device comprises: An acquisition module is used to acquire the original public opinion data corresponding to each product model included in the target product category; An analysis module, used to analyze the original public opinion data to obtain target public opinion data carried in the original public opinion data; An extraction module, used to perform hierarchical identification on the target public opinion data to obtain a public opinion data set corresponding to each product model, wherein the public opinion data set includes data of the product model in different dimensions; The analysis module is used to perform public opinion analysis based on the public opinion data set corresponding to each product model to obtain a public opinion analysis result.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images