Data processing method and system of a question and answer platform

Abstract

The present application belongs to the technical field of text data processing, and particularly relates to a data processing method and system of a question and answer platform. First, question and reply data from different regional sub-regions are collected and pre-processed to construct a data set to be analyzed. Then, a BERT model is used in combination with a supervised learning technique to accurately identify invalid response features in the reply data. Subsequently, an LDA topic model is used to cluster analyze question data associated with the invalid responses, extract a keyword feature set, and verify whether it meets the preset classification standard. If not, the keyword feature set is iteratively updated until it meets the standard. Finally, for the keyword feature set that meets the conditions, the highest priority problem features are determined, and the BERT-LSTM-CRF model is used to further extract the reasons and action feature items behind the problems. This method significantly improves the problem recognition and analysis capabilities of the question and answer platform, providing strong support for rapid response and high-quality service.

Description

Technical Field

[0001] This invention belongs to the field of text data processing technology, specifically relating to a data processing method and system for a question-and-answer platform. Background Technology

[0002] With the rapid development of internet technology, online question-and-answer platforms have become an important way for people to acquire knowledge and solve problems. These platforms provide the public with a convenient way to interact with information by collecting user questions and responses. In existing technologies, data processing for online question-and-answer platforms typically employs traditional text analysis methods, such as keyword extraction and sentiment analysis, to uncover user needs and feedback.

[0003] Existing technologies have the following main drawbacks when processing data from online question-and-answer platforms:

[0004] First, there is insufficient evaluation of response effectiveness, making it difficult for the platform to accurately identify which responses truly solve user problems, resulting in a poor user experience and wasted resources.

[0005] Secondly, the problem feature extraction is inaccurate. Traditional text analysis methods, such as rule-based keyword extraction, often fail to accurately capture the core information of the problem, which limits the platform's in-depth understanding of user needs and affects the effectiveness of subsequent problem governance activities.

[0006] Finally, existing technologies lack consideration for cross-regional spatial differences, ignoring the significant differences in user needs and problem characteristics across different regions. This results in the platform being unable to tailor its responses to local conditions and provide precise solutions when handling user inquiries, reducing the platform's response speed and service quality, and potentially leading to user dissatisfaction and churn. Summary of the Invention

[0007] The purpose of this invention is to provide a data processing method and system for a question-and-answer platform to solve the problems raised in the background art.

[0008] The present invention achieves the above objectives through the following technical solutions:

[0009] Firstly, the present invention proposes a data processing method for a question-and-answer platform, the method comprising:

[0010] The system obtains and preprocesses question data from different sub-regions based on geographical space division from online Q&A platforms to form a dataset to be analyzed; the question data includes at least two question features.

[0011] The BERT model is used to classify the response data in the dataset to be analyzed, and the invalid response characteristics of the responders are identified by combining the supervised learning algorithm.

[0012] The LDA topic model is used to perform topic clustering on the question data corresponding to the invalid response features to obtain a keyword feature set, and it is determined whether the keyword feature set meets the preset classification conditions.

[0013] If the classification criteria are not met, the keyword feature set is updated until they are met.

[0014] If the classification conditions are met, the question feature with the highest priority in the keyword feature set is determined, and the cause feature and action feature corresponding to the question feature are extracted from the invalid response feature based on the BERT-LSTM-CRF model.

[0015] Furthermore, the invalid response characteristics are specifically defined as follows: based on the degree to which the response data addresses the questioner's problem characteristics, the validity of the response data is divided into four categories: resolved, promised to be resolved, being processed, and unresolved. Unresolved responses are the invalid response characteristics.

[0016] Furthermore, the step of using the BERT model to classify the response data in the dataset to be analyzed, and combining it with a supervised learning algorithm to identify the invalid response characteristics of the responders, includes:

[0017] The initial training samples were obtained through manual annotation.

[0018] By comparing the coding results of the two annotators, samples with inconsistent annotations are discussed and identified.

[0019] The data is divided into training set, validation set and test set according to a preset ratio;

[0020] The BERT model was used for training and final prediction, and the classification results were evaluated using three metrics: recall, precision, and F1-score.

[0021] Furthermore, the LDA topic model is used to perform topic clustering on the question data corresponding to the invalid response features to obtain a keyword feature set, including:

[0022] The optimal number of question topics in the question data is determined based on topic perplexity.

[0023] Similar question topics are merged through manual inspection, and the final number of question topics and the invalid response characteristics corresponding to each question topic are determined to form a keyword feature set.

[0024] Furthermore, after performing topic clustering on the question data corresponding to the invalid response features using the LDA topic model, the method further includes:

[0025] Based on the spatial distribution characteristics of the keyword feature set, analyze the distribution of different problem features in different sub-regions; and,

[0026] The focus trend of questioners in different sub-regions on different question topics is assessed based on the amount of data related to different question topics in the question data.

[0027] Furthermore, determining whether the keyword feature set meets the preset classification conditions includes:

[0028] Based on the keyword feature set, at least two question topics are determined, and keywords with a predetermined importance ranking for each question topic are identified.

[0029] For each of the aforementioned question data, count the number of keywords belonging to the pre-defined position of each question topic contained in the question data;

[0030] If the question data contains at least three keywords from the same topic question with a predetermined position, then the question data is determined to be related to the topic question and meets the classification criteria; otherwise, it does not meet the classification criteria.

[0031] Furthermore, if the classification conditions are met, the question feature with the highest priority in the keyword feature set is determined, including:

[0032] Priority criteria for different question topics are determined based on the distribution of question topics in different sub-regions and the attention trends of questioners in different sub-regions to different question topics.

[0033] Based on the priority criteria, the question features with the highest priority in the keyword feature set are determined.

[0034] Furthermore, the extraction of causal and action features corresponding to the problem features from the invalid response features based on the BERT-LSTM-CRF model includes:

[0035] Features are extracted from the invalid response features, where BERT is used for text encoding, LSTM is used for sequence modeling, and CRF is used for sequence labeling, to identify and extract invalid response features, and further, causal features and action features related to the problem features are parsed from the invalid response features.

[0036] Furthermore, after extracting the cause features and action features corresponding to the problem features from the invalid response features based on the BERT-LSTM-CRF model, the method further includes:

[0037] Based on BETR, the causal and action features are represented as vectors. The extracted causal and action features are clustered using the K-Means model. Cosine similarity is used to measure the distance between two points. The optimal number of topics is determined by judging the sum of the distances from all points in each category to the cluster center. After further manual merging and filtering, the final causal and action features are obtained.

[0038] Based on the clustering results and the preset evaluation criteria, the importance of each causal feature is assessed, and the priority order of governance activities corresponding to different problem themes is determined based on the importance level.

[0039] Secondly, the present invention proposes a data processing system for a question-and-answer platform, the system comprising:

[0040] The data acquisition module is used to acquire and preprocess question data from questioners and response data from responders in different sub-regions based on geographical space division from the online question and answer platform to form a dataset to be analyzed; wherein, the question data includes at least two question features;

[0041] The first analysis module is used to classify the response data in the dataset to be analyzed using the BERT model, and to identify the invalid response characteristics of the respondent by combining the supervised learning algorithm.

[0042] The second analysis module is used to perform topic clustering on the question data corresponding to the invalid response features using the LDA topic model, to obtain a keyword feature set, and to determine whether the keyword feature set meets the preset classification conditions.

[0043] The data update module is used to update the keyword feature set until it meets the classification conditions when the conditions are not met.

[0044] The data processing module is used to determine the highest priority problem feature in the keyword feature set when the classification conditions are met, and extract the cause feature and action feature corresponding to the problem feature from the invalid response feature based on the BERT-LSTM-CRF model.

[0045] The beneficial effects of this invention are as follows:

[0046] 1. This invention introduces the BERT model to classify response data from online question-and-answer platforms, effectively identifying invalid response characteristics. It then performs topic clustering on the question data, extracting the highest-priority question features. In this process, the deep learning capabilities of the BERT model ensure the accuracy and efficiency of response classification, enabling the platform to quickly identify and resolve user issues. Simultaneously, accurate identification of invalid responses helps the platform intervene promptly, reducing interference from invalid responses and improving the overall effectiveness of responses and user satisfaction. Furthermore, by extracting the highest-priority question features, the platform can more accurately pinpoint user needs, providing more personalized services and further enhancing the user experience.

[0047] 2. This invention utilizes the BERT-LSTM-CRF model to extract causal and behavioral features from invalid responses, providing rich data support for the platform. Through analysis and evaluation of these features, the platform can gain a deeper understanding of the root causes and influencing factors of user problems, thereby developing more precise and effective problem management strategies. Furthermore, considering cross-regional spatial differences, the platform can develop differentiated service plans tailored to the needs and problem characteristics of users in different regions, achieving optimized resource allocation and efficient utilization. This not only improves the platform's response speed and service quality but also helps reduce operating costs and achieve sustainable development. Attached Figure Description

[0048] Figure 1 A flowchart illustrating a data processing method for a question-and-answer platform provided in this application embodiment;

[0049] Figure 2 This is a flowchart illustrating the implementation of classification conditions in the embodiments of this application;

[0050] Figure 3 This is a technical roadmap for a data processing system of a question-and-answer platform provided in an embodiment of this application. Detailed Implementation

[0051] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0052] Example 1

[0053] like Figure 1 As shown in the figure, this embodiment proposes a data processing method for a question-and-answer platform, the method including the following steps:

[0054] S1. Obtain question data from questioners and response data from responders in different sub-regions based on geographical space from an online question-and-answer platform, and preprocess them to form a dataset to be analyzed; wherein, the question data includes at least two question features; the preprocessing includes: removing special symbols and stop words from the question data and response data, and performing word segmentation to form a dataset to be analyzed.

[0055] S2. The BERT model (Bidirectional Encoder Representations from Transformers is a pre-trained model based on Transformers, mainly used for language representation) is used to classify the response data in the dataset to be analyzed, and supervised learning algorithms are combined to identify the invalid response features of the respondent. The invalid response features are specifically: the validity of the response data is divided into four categories according to the degree to which the response data solves the characteristics of the questioner's question: solved problem, promised to solve problem, problem being processed, and unsolved problem. Unsolved responses are invalid response features.

[0056] S3. Use the LDA (Latent Dirichlet Allocation) topic model to perform topic clustering on the question data corresponding to invalid response features, obtain the keyword feature set, and determine whether the keyword feature set meets the preset classification conditions;

[0057] S4. If the classification criteria are not met, update the keyword feature set until they are met;

[0058] S5. If the classification conditions are met, the question feature with the highest priority in the keyword feature set is determined, and the cause feature and action feature of the corresponding question feature are extracted from the invalid response feature based on the BERT-LSTM-CRF model.

[0059] Understandably, this embodiment achieves in-depth data mining and analysis of online Q&A platform data through five core steps. Step 1, data acquisition and preprocessing, ensures the accuracy and usability of the data, laying a solid foundation for subsequent analysis. Step 2, using the BERT model to classify the response data, accurately identifies invalid responses, effectively improving the effectiveness of the platform's responses and user satisfaction. Step 3, using the LDA topic model to perform topic clustering on the question data corresponding to invalid responses, revealing the hot issues that users are concerned about, providing valuable market insights for the platform. Step 4, determining and updating the keyword feature set, ensures the accuracy and timeliness of the analysis results, providing strong support for platform decision-making. Step 5, using the BERT-LSTM-CRF model to extract the causes and action items of problem features, deeply analyzes the essence of user problems, and provides a scientific basis for the platform to optimize services and improve user experience.

[0060] It should be noted that the BERT-LSTM-CRF model of this invention is applied to the extraction of concept phrases. This model uses BERT to encode the text to obtain a vector representation of the text, which is then input into the LTSMCRF layer to predict the label of each character in the sentence. In addition, a comparative analysis was conducted using common NER models such as Hidden Markov Model, CRF, Bi-LSTM (Bi-LSTM), and Bi-LSTM-CRF.

[0061] In a further preferred embodiment, step 2 uses the BERT model to classify the response data in the dataset to be analyzed, and combines it with a supervised learning algorithm to identify invalid response features of the responders. This includes: manually labeling to obtain the initial training samples; comparing the coding results of two labelers and discussing and determining the samples with inconsistent labels; dividing the data into training set, validation set and test set according to a preset ratio; training the BERT model and making the final prediction; and using three metrics, recall, precision and F1-score, to evaluate the classification results.

[0062] In a further preferred embodiment, step 3 uses the LDA topic model to perform topic clustering on the question data corresponding to the invalid response features to obtain a keyword feature set, including: determining the optimal number of question topics in the question data based on topic perplexity; merging similar question topics through manual inspection, and determining the final number of question topics and the invalid response features corresponding to each question topic to form a keyword feature set.

[0063] A further preferred embodiment, after performing topic clustering on the question data corresponding to invalid response features using the LDA topic model, further includes: analyzing the distribution of different question features in different sub-regions based on the spatial distribution characteristics of keyword feature sets; and assessing the attention trends of questioners in different sub-regions to different question topics based on the amount of data in the question data related to different question topics in different sub-regions.

[0064] After performing topic clustering, analyzing the distribution of different problem characteristics across different sub-regions clearly reveals the hot topics of concern to users in each region and their differences. This helps the platform understand the impact of regional culture, socioeconomic factors, and other elements on user issues, thereby enabling the development of more targeted service strategies.

[0065] like Figure 2As shown, a further preferred method is to determine whether the keyword feature set meets the preset classification conditions, including: determining at least two question topics and keywords with a priority ranking of each question topic based on the keyword feature set; for each question data, counting the number of keywords belonging to each question topic with a priority ranking in the question data; if the question data contains at least three keywords from the same question topic with a priority ranking in the question, then the question data is determined to be related to the question topic and meets the classification conditions; otherwise, it does not meet the classification conditions.

[0066] Further preferred, if the classification conditions are met, the question features with the highest priority in the keyword feature set are determined, including: determining the priority criteria for different question topics based on the distribution of question topics in different sub-regions and the attention trends of questioners in different sub-regions to different question topics; and determining the question features with the highest priority in the keyword feature set based on the priority criteria.

[0067] By employing refined keyword feature set analysis, the accuracy and efficiency of data classification are significantly improved. First, based on the keyword feature set, at least two question topics are identified, and the top-ranking keywords within each topic are selected. Then, for each question, the system counts the number of the top N keywords (N being a preset value) belonging to each question topic. If a question contains at least three of the top N keywords from the same question topic, it is considered closely related to the corresponding question topic and thus meets the classification criteria. This method not only effectively avoids the ambiguity and uncertainty of data classification but also ensures the objectivity and reliability of the classification results. Through this technology, the platform can more accurately grasp the core of user questions, providing a solid foundation for subsequent data analysis and service optimization.

[0068] A further preferred embodiment involves extracting causal and action features corresponding to the problem features from invalid response features based on the BERT-LSTM-CRF model. This includes: extracting features from invalid response features, where BERT is used for text encoding, LSTM for sequence modeling, and CRF for sequence labeling, to identify and extract invalid response features, and further parsing causal and action features related to the problem features from the invalid response features.

[0069] More specifically, causal characteristics: These characteristics aim to reveal the root causes of invalid responses. They may involve unclear question content, missing information from the questioner, insufficient knowledge or skills of the respondent, or limitations of platform rules and mechanisms. By deeply analyzing causal characteristics, the platform can identify the root cause of the problem and take targeted measures to improve it.

[0070] Correspondingly, action features, in contrast to causal features, refer to the specific actions needed to solve the problem. These actions might include providing more detailed or accurate answers, guiding the questioner to supplement necessary information, enhancing the respondent's professional knowledge and skills, or optimizing the platform's rules and processes. Extracting action features helps the platform clarify the direction for problem-solving and formulate practical action plans.

[0071] In a further preferred embodiment, after extracting the causal and action features corresponding to the problem features from the invalid response features using the BERT-LSTM-CRF model, the method further includes: representing the causal and action features as vectors based on BETR; clustering the extracted causal and action features using the K-Means model; measuring the distance between two points using cosine similarity; determining the optimal number of topics by judging the sum of the distances from all points in each category to the cluster center; and obtaining the final causal and action features after further manual merging and filtering. Based on the clustering results and preset evaluation criteria, the importance of each causal feature is evaluated, and the priority order of governance activities corresponding to different problem topics is determined based on the importance.

[0072] This embodiment demonstrates powerful data processing and analysis capabilities by integrating the BERT-LSTM-CRF model and the K-Means clustering algorithm. It first uses the BERT-LSTM-CRF model to accurately extract causal and action features from invalid responses, then uses BETR for vector representation, enhancing the computability of the features. K-Means clustering, combined with cosine similarity measurement, effectively distinguishes different features, and determines the optimal number of topics by optimizing the sum of the distances between cluster centroids, ensuring the accuracy and efficiency of clustering. Manual merging and filtering further improve the accuracy of the results. Finally, the importance of each causal feature is evaluated based on the clustering results and preset criteria, providing a scientific basis for prioritizing problem-related governance activities. This helps the platform respond quickly and resolve user issues, improving service quality and user experience.

[0073] According to the above embodiments of the present invention, the data processing method, in specific implementation, includes the following steps:

[0074] ① Data acquisition and preprocessing

[0075] First, question data from users and response data from users in different geographically segmented sub-regions were obtained from an online Q&A platform. This data covers detailed information about user questions and the content of user responses. To ensure the accuracy and usability of the data, preprocessing was performed, including removing special symbols and stop words from the question and response data, and performing word segmentation to form the dataset to be analyzed. This step laid a solid foundation for subsequent data analysis.

[0076] ② Response data classification and invalid response identification

[0077] The BERT model is used to classify the response data in the dataset to be analyzed. The BERT model is a pre-trained Transformer-based model that accurately captures contextual information in text, making it well-suited for language representation. Combined with supervised learning algorithms, this invention identifies invalid response features. Specifically, based on the degree to which the response data addresses the questioner's problem features, the validity of the response data is categorized into four types: resolved, committed to resolved, currently being addressed, and unresolved. Unresolved responses are considered invalid response features. To improve classification accuracy, this invention also uses manual annotation to obtain the initial training samples, compares the annotation results, divides the data, and uses the BERT model for training and prediction. The classification results are evaluated using three metrics: recall, precision, and F1-score.

[0078] ③ Question data topic clustering and keyword feature set determination

[0079] For question data corresponding to invalid response characteristics, LDA topic modeling is used for topic clustering. LDA topic modeling can reveal the underlying topic structure in text, making it very suitable for mining hot topics of interest to users. This invention determines the optimal number of question topics in the question data based on topic perplexity and merges similar topics through manual inspection to form a keyword feature set. Furthermore, it analyzes the distribution of different question features in different sub-regions based on the spatial distribution characteristics of the keyword feature set, and assesses the attention trends of questioners in different sub-regions to different question topics.

[0080] To determine whether a keyword feature set meets preset classification criteria, this invention identifies the question topic and its top-ranking keywords based on the keyword feature set. For each question, the number of the top N keywords belonging to each question topic (N is a preset value) is counted. If a question contains at least three of the top N keywords from the same question topic, it is determined that the question is relevant to the question topic and meets the classification criteria.

[0081] ④ Keyword feature set update

[0082] If the keyword feature set does not meet the classification criteria, an update operation is performed. Specifically, based on the new question and response data, topic clustering is re-performed, and the keyword feature set is updated until it meets the classification criteria. This step ensures the accuracy and timeliness of the analysis results.

[0083] ⑤ Problem feature extraction and identification of cause and action features

[0084] After the keyword feature set meets the classification criteria, the question feature with the highest priority is determined. This invention determines the priority criteria for different question topics based on the distribution of question topics in different sub-regions and the questioner's attention trends to different question topics, and then determines the question feature with the highest priority in the keyword feature set based on these priority criteria.

[0085] Next, the BERT-LSTM-CRF model is used to extract causal and action features corresponding to the problem features from the invalid response features. BERT is used for text encoding, LSTM for sequence modeling, and CRF for sequence labeling. This combination can accurately identify and extract key information from the invalid response features. Furthermore, causal and action features related to the problem features are parsed from the invalid response features. To evaluate the importance of each causal feature, the K-Means model is used to cluster the extracted causal and action features, and the priority order of governance activities corresponding to different problem topics is determined based on the clustering results and preset evaluation criteria.

[0086] like Figure 3 As shown, based on the same inventive concept, this invention also proposes a data processing system for a question-and-answer platform, the system comprising:

[0087] The data acquisition module is used to acquire and preprocess question data from questioners and response data from responders in different sub-regions based on geographical space division from the online question and answer platform to form a dataset to be analyzed; wherein, the question data includes at least two question features;

[0088] The first analysis module is used to classify the response data in the dataset to be analyzed using the BERT model, and to identify the invalid response characteristics of the respondent by combining supervised learning algorithms.

[0089] The second analysis module is used to perform topic clustering on the question data corresponding to invalid response features using the LDA topic model, obtain keyword feature sets, and determine whether the keyword feature sets meet the preset classification conditions.

[0090] The classification criteria are as follows: Based on the keyword feature set, at least two question topics are determined, and the keywords with the highest importance ranking for each question topic are identified; for each question data, the number of keywords belonging to the highest importance ranking for each question topic is counted; if the question data contains at least three keywords from the same question topic with the highest importance ranking, the question data is determined to be related to the question topic and meets the classification criteria; otherwise, it does not meet the classification criteria.

[0091] The data update module is used to update the keyword feature set until it meets the classification criteria when the criteria are not met.

[0092] The data processing module is used to determine the highest priority question features in the keyword feature set when the classification conditions are met, and extract the cause features and action features of the corresponding question features from the invalid response features based on the BERT-LSTM-CRF model.

[0093] It should be noted that each module in the above data processing system corresponds to steps S1 to S5 in implementing the above data processing method. The instances and application scenarios implemented by multiple modules and their corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above.

[0094] Understandably, the aforementioned data processing system can be built into an online question-and-answer platform. The goal of constructing such an efficient data processing system is to optimize the quality of questions and answers and the user experience on the online question-and-answer platform.

[0095] In another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the steps of any of the above data processing methods.

[0096] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the steps of any of the data processing methods described above.

[0097] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated.

[0098] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0101] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data processing method for a question-and-answer platform, characterized in that, The method includes: The system obtains and preprocesses question data from different sub-regions based on geographical space division from online Q&A platforms to form a dataset to be analyzed; the question data includes at least two question features. The BERT model is used to classify the response data in the dataset to be analyzed, and the invalid response characteristics of the responders are identified by combining the supervised learning algorithm. Based on the degree to which the response data addresses the characteristics of the questioner's problem, the validity of the response data is divided into four categories: resolved, promised to be resolved, being processed, and unresolved. Unresolved responses are the invalid response characteristics. The LDA topic model is used to perform topic clustering on the question data corresponding to the invalid response features to obtain a keyword feature set, and it is determined whether the keyword feature set meets the preset classification conditions. After performing topic clustering on the question data corresponding to the invalid response features using the LDA topic model, the method further includes: Based on the spatial distribution characteristics of the keyword feature set, analyze the distribution of different problem features in different sub-regions; and, The focus trend of questioners in different sub-regions on different question topics is assessed based on the amount of data related to different question topics in the question data. The step of determining whether the keyword feature set meets the preset classification conditions includes: Based on the keyword feature set, at least two question topics are determined, and keywords for each question topic are ranked first in importance according to a predetermined position; For each of the aforementioned question data, count the number of keywords belonging to the pre-defined position of each question topic contained in the question data; If the question data contains at least three keywords from the same topic question with a predetermined ranking, then the question data is determined to be related to the topic question and meets the classification criteria; otherwise, it does not meet the classification criteria. If the classification criteria are not met, the keyword feature set is updated until they are met. If the classification conditions are met, the question feature with the highest priority in the keyword feature set is determined, and the cause feature and action feature corresponding to the question feature are extracted from the invalid response feature based on the BERT-LSTM-CRF model. If the classification conditions are met, the question features with the highest priority in the keyword feature set are determined, including: Priority criteria for different question topics are determined based on the distribution of question topics in different sub-regions and the attention trends of questioners in different sub-regions to different question topics. Based on the priority criteria, the question features with the highest priority in the keyword feature set are determined.

2. The data processing method for a question-and-answer platform according to claim 1, characterized in that: The step of using the BERT model to classify the response data in the dataset to be analyzed, and combining it with a supervised learning algorithm to identify the characteristics of invalid responses from the responders, includes: The initial training samples were obtained through manual annotation. By comparing the coding results of the two annotators, samples with inconsistent annotations are discussed and identified. The data is divided into training set, validation set and test set according to a preset ratio; The BERT model was used for training and final prediction, and the classification results were evaluated using three metrics: recall, precision, and F1-score.

3. The data processing method for a question-and-answer platform according to claim 1, characterized in that: The LDA topic model is used to perform topic clustering on the question data corresponding to the invalid response features to obtain a keyword feature set, including: The optimal number of question topics in the question data is determined based on topic perplexity. Similar question topics are merged through manual inspection, and the final number of question topics and the invalid response characteristics corresponding to each question topic are determined to form a keyword feature set.

4. The data processing method for a question-and-answer platform according to claim 1, characterized in that: The BERT-LSTM-CRF model extracts causal and action features corresponding to the problem features from the invalid response features, including: Features are extracted from the invalid response features, where BERT is used for text encoding, LSTM is used for sequence modeling, and CRF is used for sequence labeling, to identify and extract invalid response features, and further, causal features and action features related to the problem features are parsed from the invalid response features.

5. The data processing method for a question-and-answer platform according to claim 1, characterized in that: After extracting the causal and action features corresponding to the problem features from the invalid response features using the BERT-LSTM-CRF model, the method further includes: Based on BETR, the causal and action features are represented as vectors. The extracted causal and action features are clustered using the K-Means model. Cosine similarity is used to measure the distance between two points. The optimal number of topics is determined by judging the sum of the distances from all points in each category to the cluster center. After further manual merging and filtering, the final causal and action features are obtained. Based on the clustering results and the preset evaluation criteria, the importance of each causal feature is assessed, and the priority order of governance activities corresponding to different problem themes is determined based on the importance level.

6. A data processing system for a question-and-answer platform, characterized in that: The system includes: The data acquisition module is used to acquire and preprocess question data from questioners and response data from responders in different sub-regions based on geographical space division from the online question and answer platform to form a dataset to be analyzed; wherein, the question data includes at least two question features; The first analysis module is used to classify the response data in the dataset to be analyzed using the BERT model, and to identify the invalid response characteristics of the respondent by combining the supervised learning algorithm. Based on the degree to which the response data addresses the characteristics of the questioner's problem, the validity of the response data is divided into four categories: resolved, promised to be resolved, being processed, and unresolved. Unresolved responses are the invalid response characteristics. The second analysis module is used to perform topic clustering on the question data corresponding to the invalid response features using the LDA topic model, to obtain a keyword feature set, and to determine whether the keyword feature set meets the preset classification conditions. After performing topic clustering on the question data corresponding to the invalid response features using the LDA topic model, the method further includes: Based on the spatial distribution characteristics of the keyword feature set, analyze the distribution of different problem features in different sub-regions; and, The focus trend of questioners in different sub-regions on different question topics is assessed based on the amount of data related to different question topics in the question data. The step of determining whether the keyword feature set meets the preset classification conditions includes: Based on the keyword feature set, at least two question topics are determined, and keywords for each question topic are ranked first in importance according to a predetermined position; For each of the aforementioned question data, count the number of keywords belonging to the pre-defined position of each question topic contained in the question data; If the question data contains at least three keywords from the same topic question with a predetermined ranking, then the question data is determined to be related to the topic question and meets the classification criteria; otherwise, it does not meet the classification criteria. The data update module is used to update the keyword feature set until it meets the classification conditions when the conditions are not met. The data processing module is used to determine the highest priority question feature in the keyword feature set when the classification conditions are met, and extract the cause feature and action feature corresponding to the question feature from the invalid response feature based on the BERT-LSTM-CRF model; If the classification conditions are met, the question features with the highest priority in the keyword feature set are determined, including: Priority criteria for different question topics are determined based on the distribution of question topics in different sub-regions and the attention trends of questioners in different sub-regions to different question topics. Based on the priority criteria, the question features with the highest priority in the keyword feature set are determined.

Images