Public investigation method and system based on big data analysis and AI technology

Abstract

The invention discloses a public investigation method and system based on big data analysis and an AI technology, belongs to the technical field of data investigation, is used for public investigation, and comprises the steps of obtaining public information, performing preliminary classification on the public information, performing keyword capturing on a result after preliminary classification, sorting the captured keywords, and combining a keyword sorting result to obtain a public investigation result. Obtaining a civil investigation result; inputting the public information and the public investigation results for multiple times into a deep learning model for model training, and inputting the collected public information into the deep learning model after the model training is completed to obtain the public investigation results output by the deep learning model. According to the invention, social management is enabled by big data, decision making is more scientific and effective, social governance is improved in quality and efficiency, and service economic and social development and livelihood business improvement are accurately assisted.

Description

Technical Field

[0001] The present invention discloses a method and system for public opinion survey based on big data analysis and AI technology, belonging to the field of data survey technology. Background Art

[0002] Understanding public opinion is a necessary means of enhancing government services and a necessary path to understanding the needs of the masses. Public opinion surveys are a type of social survey designed to understand public opinion trends. They utilize scientific survey and statistical methods to truthfully reflect the attitudes of a certain range of people towards a particular social issue or issues. In terms of content, they fall within the scope of public opinion surveys; in terms of methodology, they fall within the scope of sampling surveys. Existing methods for public opinion surveys primarily rely on questionnaires and manual statistics, which are slow and prone to omissions and omissions in manual screening, failing to fully reflect public opinion needs. Simply collecting public opinion without screening and classification results in an excessively large amount of collected public opinion, making it difficult to support decision-making. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for public opinion surveys based on big data analysis and AI technology to solve the problem in the prior art that public opinion surveys rely on manual labor and thus have low efficiency.

[0004] The public opinion survey method based on big data analysis and AI technology includes obtaining public opinion information, preliminarily classifying the public opinion information, capturing keywords for the results of the preliminary classification, organizing the captured keywords, and obtaining public opinion survey results based on the keyword organization results; inputting multiple public opinion information and public opinion survey results into a deep learning model for model training. After the model training is completed, the collected public opinion information is input into the deep learning model to obtain the public opinion survey results output by the deep learning model.

[0005] Obtaining public opinion information includes using big data analysis software to collect government question and answer records from government service websites and citizen hotline recording platforms.

[0006] The preliminary classification includes dividing public opinion information into two levels of classification. The first level classification includes dividing public opinion information into answered information and unanswered information. The second level classification includes dividing answered information into information with follow-up questions and information without follow-up questions. The information with follow-up questions is public opinion information with additional consultation content after the reply, and the information without follow-up questions is public opinion information without additional consultation content after the reply.

[0007] The keyword capture includes capturing the submission time of the unanswered information, and sequentially capturing nouns and verbs for the information without follow-up questions.

[0008] Sorting out the captured keywords includes recording unanswered information whose submission time exceeds the preset time threshold, arranging the captured nouns from high to low according to the number, filtering according to the preset number of nouns, and eliminating nouns with less than the preset number of nouns. For each filtered noun, arranging the verbs from high to low according to the number of captured verbs, filtering according to the preset number of verbs, and eliminating verbs with less than the preset number of verbs.

[0009] The submission time of the unanswered information is screened, and submission times later than a preset time are eliminated. The preset time, the preset number of nouns, and the preset number of verbs are determined by a large language model using AI technology.

[0010] The opinion poll results include unanswered information with a submission time later than a preset time, information without follow-up questions corresponding to verbs with a number greater than a preset number of verbs, and information with follow-up questions.

[0011] The sequence of unanswered information whose submission time is later than the preset time is α = [content of unanswered information, submission time of unanswered information];

[0012] The sequence with follow-up information is β = [content of the follow-up case];

[0013] The sequence of answered information corresponding to verbs with a number higher than the preset verb number is [X1, X2......X n ],X1=[x1、x2......x m ] T , X n is the noun of the nth unanswered message, x m It's X n The mth verb in .

[0014] The deep learning model includes three feature processing branches, the first branch input is α, the second branch input is β, and the third branch input is [X1, X2...X n ];

[0015] The first branch passes through 6 convolution blocks in series, and then performs average pooling to generate the first branch output;

[0016] The second branch passes through 6 convolution blocks in series, and then performs average pooling to generate the second branch output;

[0017] The third branch passes through 6 series convolution blocks in sequence to generate the output result. The features before each convolution block are average pooled to obtain 6 average pooling results. The output result and the 6 average pooling results are fused to generate the output of the third branch.

[0018] The output results of the three branches are feature-concatenated and the poll results are output after passing through 6 fully connected layers.

[0019] The public opinion survey system based on big data analysis and AI technology uses the public opinion survey method based on big data analysis and AI technology, including a public opinion information acquisition module, a preliminary classification module, a keyword capture module, a deep learning model module, and a public opinion survey result output module.

[0020] Compared with the existing technology, the present invention has the following beneficial effects: the present invention empowers social management with big data, realizes more scientific and effective decision-making and improves the quality and efficiency of social governance, and accurately helps serve economic and social development and improves people's livelihood; integrates into local government government websites, citizen hotlines and other data platforms, and uses big data analysis and AI artificial intelligence technology to sort out and extract issues that are of high public concern and concentrated bottlenecks and pain points from daily information, sort them, and create hot lists, timely collect public voices from all walks of life, screen hot public opinions, widely understand public needs, and respond to public wishes in a timely manner. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] The public opinion survey method based on big data analysis and AI technology includes obtaining public opinion information, preliminarily classifying the public opinion information, capturing keywords for the results of the preliminary classification, organizing the captured keywords, and obtaining public opinion survey results based on the keyword organization results; inputting multiple public opinion information and public opinion survey results into a deep learning model for model training. After the model training is completed, the collected public opinion information is input into the deep learning model to obtain the public opinion survey results output by the deep learning model.

[0023] Obtaining public opinion information includes using big data analysis software to collect government question and answer records from government service websites and citizen hotline recording platforms.

[0024] The preliminary classification includes dividing public opinion information into two levels of classification. The first level classification includes dividing public opinion information into answered information and unanswered information. The second level classification includes dividing answered information into information with follow-up questions and information without follow-up questions. The information with follow-up questions is public opinion information with additional consultation content after the reply, and the information without follow-up questions is public opinion information without additional consultation content after the reply.

[0025] The keyword capture includes capturing the submission time of the unanswered information, and sequentially capturing nouns and verbs for the information without follow-up questions.

[0026] Sorting out the captured keywords includes recording unanswered information whose submission time exceeds the preset time threshold, arranging the captured nouns from high to low according to the number, filtering according to the preset number of nouns, and eliminating nouns with less than the preset number of nouns. For each filtered noun, arranging the verbs from high to low according to the number of captured verbs, filtering according to the preset number of verbs, and eliminating verbs with less than the preset number of verbs.

[0027] The submission time of the unanswered information is screened, and submission times later than a preset time are eliminated. The preset time, the preset number of nouns, and the preset number of verbs are determined by a large language model using AI technology.

[0028] The opinion poll results include unanswered information with a submission time later than a preset time, information without follow-up questions corresponding to verbs with a number greater than a preset number of verbs, and information with follow-up questions.

[0029] The sequence of unanswered information whose submission time is later than the preset time is α = [content of unanswered information, submission time of unanswered information];

[0030] The sequence with follow-up information is β = [content of the follow-up case];

[0031] The sequence of answered information corresponding to verbs with a number higher than the preset verb number is [X1, X2......X n ],X1=[x1、x2......x m ] T , X n is the noun of the nth unanswered message, x m It's X n The mth verb in .

[0032] The deep learning model includes three feature processing branches, the first branch input is α, the second branch input is β, and the third branch input is [X1, X2...X n ];

[0033] The first branch passes through 6 convolution blocks in series, and then performs average pooling to generate the first branch output;

[0034] The second branch passes through 6 convolution blocks in series, and then performs average pooling to generate the second branch output;

[0035] The third branch passes through 6 series convolution blocks in sequence to generate the output result. The features before each convolution block are average pooled to obtain 6 average pooling results. The output result and the 6 average pooling results are fused to generate the output of the third branch.

[0036] The output results of the three branches are feature-concatenated and the poll results are output after passing through 6 fully connected layers.

[0037] The public opinion survey system based on big data analysis and AI technology uses the public opinion survey method based on big data analysis and AI technology, including a public opinion information acquisition module, a preliminary classification module, a keyword capture module, a deep learning model module, and a public opinion survey result output module.

[0038] The present invention verifies the above technical solution through an embodiment, and takes a provincial government service network as an example for implementation. The government service network discloses public opinion information in the service consultation, as shown in Table 1.

[0039] Table 1. Public Opinion Information

[0040]

[0041] Table 1 shows the title, handling department, submission date, reply status, and follow-up status. Keywords are captured based on the information in Table 1.

[0042] According to the handling departments, taking the XX City Natural Resources and Planning Bureau as an example, the noun capture results are obtained, as shown in Table 2.

[0043] Table 2. Noun crawling results of XX City Natural Resources and Planning Bureau

[0044] noun Handling Department Submission Date Reply Status Question status rural residential land XX City Natural Resources and Planning Bureau 2021-01-25 Replied No follow-up questions Real estate information XX Municipal Natural Resources and Planning Bureau 2020-12-30 Replied No follow-up questions Commercial housing XX City Natural Resources and Planning Bureau 2020-12-24 Replied No follow-up questions real estate XX Municipal Natural Resources and Planning Bureau 2020-12-22 Replied No follow-up questions parking space XX Municipal Natural Resources and Planning Bureau 2020-12-09 Replied No follow-up questions Abandoned mines XX Municipal Natural Resources and Planning Bureau 2020-08-10 Replied No follow-up questions approval document XX Municipal Natural Resources and Planning Bureau 2020-05-01 Replied No follow-up questions Co-ownership certificate XX City Natural Resources and Planning Bureau 2020-04-25 Replied No follow-up questions Golf Course XX Municipal Natural Resources and Planning Bureau 2019-04-29 Replied No follow-up questions arable land XX Municipal Natural Resources and Planning Bureau 2019-04-15 Replied No follow-up questions Power generation projects XX Municipal Natural Resources and Planning Bureau 2018-12-24 Replied No follow-up questions Newborns XX Municipal Natural Resources and Planning Bureau 2018-12-18 Replied No follow-up questions Urban and rural construction land XX Municipal Natural Resources and Planning Bureau 2018-12-06 Replied No follow-up questions land XX Municipal Natural Resources and Planning Bureau 2018-08-02 Replied No follow-up questions Real estate certificate XX Municipal Natural Resources and Planning Bureau 2018-04-26 Replied No follow-up questions shantytowns XX Municipal Natural Resources and Planning Bureau 2018-01-23 Replied No follow-up questions ;

[0045] In the deep learning model, the first and second branches used in the present invention respectively perform feature extraction of unanswered information with a submission time higher than the preset time and sequences with follow-up information. In order to achieve effective extraction of names and verbs in semantics, the present invention designs a feature branch before the convolution block in the third branch to better extract verbs and names.

[0046] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. The public opinion survey method based on big data analysis and AI technology is characterized by: The method includes obtaining public opinion information, preliminarily classifying the public opinion information, capturing keywords for the results of the preliminary classification, organizing the captured keywords, and obtaining public opinion survey results based on the keyword organization results; inputting multiple public opinion information and public opinion survey results into a deep learning model for model training. After the model training is completed, the collected public opinion information is input into the deep learning model to obtain the public opinion survey results output by the deep learning model.

2. The public opinion survey method based on big data analysis and AI technology according to claim 1 is characterized in that: Obtaining public opinion information includes using big data analysis software to collect government question and answer records from government service websites and citizen hotline recording platforms.

3. The public opinion survey method based on big data analysis and AI technology according to claim 2 is characterized in that: The preliminary classification includes dividing public opinion information into two levels of classification. The first level classification includes dividing public opinion information into answered information and unanswered information. The second level classification includes dividing answered information into information with follow-up questions and information without follow-up questions. The information with follow-up questions is public opinion information with additional consultation content after the reply, and the information without follow-up questions is public opinion information without additional consultation content after the reply.

4. The public opinion survey method based on big data analysis and AI technology according to claim 3 is characterized in that: The keyword capture includes capturing the submission time of the unanswered information, and sequentially capturing nouns and verbs for the information without follow-up questions.

5. The public opinion survey method based on big data analysis and AI technology according to claim 4 is characterized in that: Sorting out the captured keywords includes recording unanswered information whose submission time exceeds the preset time threshold, arranging the captured nouns from high to low according to the number, filtering according to the preset number of nouns, and eliminating nouns with less than the preset number of nouns. For each filtered noun, arranging the verbs from high to low according to the number of captured verbs, filtering according to the preset number of verbs, and eliminating verbs with less than the preset number of verbs.

6. The public opinion survey method based on big data analysis and AI technology according to claim 5 is characterized in that: The submission time of the unanswered information is screened, and submission times later than a preset time are eliminated. The preset time, the preset number of nouns, and the preset number of verbs are determined by a large language model using AI technology.

7. The public opinion survey method based on big data analysis and AI technology according to claim 6 is characterized in that: The opinion poll results include unanswered information with a submission time later than a preset time, information without follow-up questions corresponding to verbs with a number greater than a preset number of verbs, and information with follow-up questions.

8. The public opinion survey method based on big data analysis and AI technology according to claim 7 is characterized in that: The sequence of unanswered information whose submission time is later than the preset time is α = [content of unanswered information, submission time of unanswered information]; The sequence with follow-up information is β = [content of the follow-up case]; The sequence of answered information corresponding to verbs with a number higher than the preset verb number is [X1, X2......X n ],X1=[x1、x2......x m ] T , X n is the noun of the nth unanswered message, x m It's X n The mth verb in .

9. The public opinion survey method based on big data analysis and AI technology according to claim 8, characterized in that: The deep learning model includes three feature processing branches, the first branch input is α, the second branch input is β, and the third branch input is [X1, X2...X n ]; The first branch passes through 6 convolution blocks in series, and then performs average pooling to generate the first branch output; The second branch passes through 6 convolution blocks in series, and then performs average pooling to generate the second branch output; The third branch passes through 6 series-connected convolution blocks in sequence to produce the output result. The features before each convolution block are average pooled to obtain 6 average pooling results. The output result and the 6 average pooling results are fused to produce the output of the third branch; the output results of the three branches are feature-concatenated and the poll results are output after passing through 6 fully connected layers.

10. The public opinion survey system based on big data analysis and AI technology is characterized by: The public opinion survey method based on big data analysis and AI technology according to claim 9 includes a public opinion information acquisition module, a preliminary classification module, a keyword capture module, a deep learning model module, and a public opinion survey result output module.

Images