Government affair text quality evaluation method based on semantic analysis and big data index

Abstract

The invention provides a semantic analysis and big data index-based government affair text quality evaluation method. The method comprises the following steps of 1, setting a benchmarking data acquisition source according to n dimensions needing to be embodied by a to-be-evaluated government affair text, and performing data acquisition; 2, text data preprocessing: performing text data preprocessing on the n benchmarking text data sets acquired in the first step; 3, performing natural language processing on the standard government affair text data set and the to-be-evaluated government affair text data set; and 4, calculating n sub-indexes. The cosine similarity between the to-be-evaluated government affair text data set and the n benchmarking government affair text data sets is calculated in sequence, and n score indexes rho score are obtained in sequence; and 5, weighting the n sub-indexes to obtain a total index rho total. According to the method, the big data index for evaluating the quality of the government affair text is constructed by using big data and an artificial intelligence technology, and an objective and intelligent quality evaluation result can be provided for the quality of the government affair text from a correlation angle through an intelligent algorithm.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a government text quality evaluation method based on semantic analysis and big data index. Background Art

[0002] The quality of government texts is one of the core elements that reflect the performance of state employees, and it is also the most difficult part to quantify and assess. At present, in the existing technology, the quality of government texts generally adopts a subjective evaluation method, that is, it is carried out through manual evaluation, which mainly relies on human judgment, including the clarity of the theme and purpose: evaluate whether the government text clearly expresses the theme and purpose so that readers can quickly understand and grasp the key information. Reasonableness of logical structure: check whether the government text organizes the content according to a reasonable logical structure, including the introduction, body and ending, to ensure that the hierarchy of information is clear. Accuracy of information presentation: evaluate whether the government text accurately conveys relevant information without omissions or misleading, and ensure that the content is complete and credible. For example, experts can conduct subjective evaluation and scoring of the text to be evaluated by holding an expert review meeting, and evaluate the quality of the text in terms of readability, comprehensibility, accuracy of emotional expression, clarity of language, consistency of expression theme, logic, amount of information contained, and structure.

[0003] Manual evaluation can use techniques such as checking information accuracy, conciseness, logical coherence, using concise language, paying attention to format specifications, and proofreading texts to conduct a detailed quality evaluation of government documents. This evaluation method has three disadvantages: first, it is not easy to operate. In order to ensure fairness, it is necessary to convene multiple experts for evaluation; second, it is easily affected by subjective factors; third, the evaluation results are not easy to quantify and compare. Therefore, it is urgent to study a method that does not rely entirely on manual methods, and can give objective and intelligent quality evaluation results for government documents from the perspective of relevance through intelligent algorithms, and can give quantitative evaluation results for the level of relevance to evaluate the quality of government documents.

[0004] At present, some technology-assisted evaluation methods have emerged:

[0005] Big data index: Big data index refers to all indexes that use big data as the research object. It complements the traditional statistical survey index and is a relative index used to reflect the comprehensive change degree of a phenomenon. The big data index combines the characteristics of big data and statistical indexes. It has similar mathematical, economic or sociological properties to traditional statistical survey indexes, and has big data properties that traditional statistical survey indexes do not have, reflecting extensiveness, continuity and refinement.

[0006] Natural language processing technology: With the development of technology, natural language processing technology (NLP) has been gradually applied to digital government application scenarios. NLP technology can automatically extract key information from texts, such as policy topics, target groups, policy measures, etc., and perform semantic analysis on texts.

[0007] The government text quality evaluation method based on semantic analysis and big data index proposed in the present invention is a combination of the above two methods, which can assist evaluators in evaluating the quality of texts.

[0008] It should be noted that the above introduction to the technical background is only for the convenience of providing a clear and complete description of the technical solutions of the present invention and for the convenience of understanding by those skilled in the art. It cannot be considered that the above technical solutions are well known to those skilled in the art simply because they are described in the technical background section of the present invention. Summary of the invention

[0009] The purpose of the present invention is to provide a government text quality evaluation method based on semantic analysis and big data index, and to construct a big data index for evaluating the quality of government texts by using big data and artificial intelligence technology. The index is composed of a "1+n" index system, and can give objective and intelligent quality evaluation results for government texts from the perspective of relevance through intelligent algorithms, and can give quantitative evaluation results for the level of relevance. Compared with traditional methods, it can reflect the comprehensive degree of change of complex factors that cannot be directly added, and can evaluate the text quality at multiple scales and dimensions and give quantitative indicators: for example, it can measure the overall change trend of government texts of different sessions; it can measure the changes and response levels of given topics; and it can make comparisons between single texts.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] The present invention provides a method for evaluating the quality of government documents based on semantic analysis and big data index, comprising the following steps:

[0012] Step 1: Set the benchmark data collection source and collect data according to the n dimensions that the government documents to be evaluated need to reflect;

[0013] Step 2: Text data preprocessing: performing text data preprocessing on the n benchmark text data sets collected in the first step, wherein the text data preprocessing includes text data denoising and deduplication;

[0014] Step 3: Natural language processing of benchmarking government text datasets and government text datasets to be evaluated;

[0015] The n benchmark text datasets obtained after the second step of text data preprocessing and the government text dataset to be evaluated are processed using natural language processing technology, including word segmentation, part-of-speech tagging, text vectorization, named entity recognition, keyword extraction, and text classification and clustering operations, to obtain n benchmark government text datasets corresponding to n sub-indexes;

[0016] Step 4: Calculation of n sub-indices;

[0017] Calculate the cosine similarity between the government text dataset to be evaluated and the n benchmark government text datasets in turn, and obtain the sub-index ρ in turn 分 ;

[0018] The cosine similarity is used to obtain the relevance between the government document to be evaluated and the n topics, which is the score index ρ 分 , the cosine similarity algorithm is as follows:

[0019]

[0020] A represents the government text vector to be evaluated, and B represents the benchmark subject vector. The closer the cosine value cosθ is to 1, the smaller the distance between the government text vector to be evaluated and the benchmark subject vector is, and the higher the degree of correlation is. Vectors A and B represent word frequency vectors. i , B i Represents the components of vectors A and B respectively, i is the dimension of vectors A and B, cosθ is the cosine similarity, and the index is ρ 分 ; The big data index is composed of a "1+n" index system, including a total index ρ 总 and n sub-indices ρ 分 ; n is a positive integer;

[0021] ρ 分 = cosθ

[0022] Step 5: Weight the n sub-indices to obtain the total index ρ 总 .

[0023] Optionally, in the first step, a benchmark data collection source is set to collect n dimensions of content that need to be reflected in the government text to be evaluated, where n=5, including: collecting target sample policies, target sample government work reports, target sample social public opinion information, target sample netizen message clues, and target sample topic information that need to be reflected;

[0024] In the third step: benchmarking government text data set construction;

[0025] The benchmark text data set obtained after the second step of text data preprocessing is processed using natural language processing technology to obtain 5 benchmark government text data sets corresponding to 5 sub-indices; define the sub-index ρ 分They are: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5; the total index ρ is obtained by weighting the five sub-indexes: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5. 总 .

[0026] In summary, the present invention uses big data and artificial intelligence technology to construct a big data index for evaluating the quality of government texts. The index consists of a "1+n" index system, including 1 total index and n sub-indexes. The big data index is calculated by judging the degree of fit between government texts and various aspects at the semantic level, thereby reflecting the response of government texts to various topics. That is, the big data index for evaluating the quality of government texts constructed by the present invention using big data and artificial intelligence technology can give objective and intelligent quality evaluation results for the quality of government texts from the perspective of relevance through intelligent algorithms.

[0027] The government document quality evaluation method based on semantic analysis and big data index provided by the present invention can give objective and intelligent quality evaluation results for government documents from the perspective of relevance through intelligent algorithms, and can give quantitative evaluation results for the degree of relevance. It is helpful to quantitatively reflect and evaluate the contribution of government documents in implementing central policies, connecting key tasks, and reflecting social conditions and public opinion. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 It is a data set representation diagram of a government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0030] Figure 2 It is a flow chart of a method for evaluating the quality of government texts based on semantic analysis and big data index according to an embodiment of the present invention.

[0031] Figure 3 It is a flow chart of the 1+n index system of the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0032] Figure 4 This is a detailed flow chart of a method for evaluating the quality of government texts based on semantic analysis and big data index according to an embodiment of the present invention.

[0033] Figure 5 It is a flow chart of the 1+5 index system in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0034] Figure 6 It is a list of proposal keywords in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0035] Figure 7 This is a graph of the correspondence between various names and the number of cases submitted in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0036] Figure 8 It is a ranking diagram of the top 10 keywords with total scores in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0037] Fig. 9 It is a ranking chart of the top 10 keywords in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention in terms of policy responsiveness, report fit, public concern, clue utilization and topic participation.

[0038] Fig.10 It is a schematic diagram of the analysis of 97 keywords and the keyword policy responsiveness index in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0039] Fig.11 It is a schematic diagram of the analysis of 97 keywords in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention, and the keyword report fit index.

[0040] Fig.12 It is a schematic diagram of the analysis of 97 keywords in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention, and the keyword public concern index.

[0041] Fig.13 It is a schematic diagram of the analysis of 97 keywords and the keyword clue utilization index in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0042] Fig.14 It is a schematic diagram of 97 keyword analyses and topic participation index in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0043] Fig.15 This is a schematic diagram of 11 major classifications of proposal keywords in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0044] Fig.16 It is a schematic diagram of the proposal topic classification index in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0045] Fig.17 It is a schematic diagram of the policy responsiveness index for proposal topic classification in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention.

[0046] Fig.18 It is a schematic diagram of the proposal subject classification report fit index in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0048] The importance of government text quality evaluation is mainly reflected in the following aspects:

[0049] Ensure the accuracy of information: Government documents involve important information such as policies and regulations, and their accuracy is directly related to the public interest and the image of the government.

[0050] Improve the readability of information: Clear and concise government texts help the public quickly understand policy content and improve the popularity and implementation efficiency of policies.

[0051] Shaping the government’s image: High-quality government documents can demonstrate the government’s professionalism and efficient execution, and enhance the public’s trust in the government.

[0052] Therefore, the quality of government texts is one of the core elements that reflect the performance of government staff, and it is also the most difficult part to quantify and assess. At present, in the existing technology, the quality of government texts generally adopts subjective evaluation methods, that is, it is carried out through manual evaluation, mainly relying on human judgment, including the clarity of the theme and purpose: evaluate whether the government text clearly expresses the theme and purpose so that readers can quickly understand and grasp the key information. Logical structure rationality: check whether the government text organizes the content according to a reasonable logical structure, including the introduction, body and ending, to ensure that the hierarchy of information is clear. Accuracy of information presentation: evaluate whether the government text accurately conveys relevant information without omissions or misleading, and ensure that the content is complete and credible. For example, experts can conduct subjective evaluation and scoring of the text to be evaluated by holding an expert review meeting, and evaluate the quality of the text in terms of readability, comprehensibility, accuracy of emotional expression, clarity of language, consistency of expression theme, logic, amount of information contained, and structure.

[0053] Manual evaluation can use techniques such as checking information accuracy, conciseness, logical coherence, using concise language, paying attention to format specifications, and proofreading texts to conduct a detailed quality assessment of government documents. This evaluation method has three disadvantages: first, it is not easy to operate. In order to ensure fairness, it is necessary to convene multiple experts for evaluation; second, it is easily affected by subjective factors; third, the evaluation results are not easy to quantify and compare. Therefore, relying solely on humans to complete it has many disadvantages and is no longer able to meet the needs of social development.

[0054] At present, there are some existing technology-assisted evaluation methods, such as natural language processing technology: With the development of technology, natural language processing technology (NLP) has been used in digital government application scenarios in small quantities. NLP technology can automatically extract key information from texts, such as policy themes, target groups, policy measures, etc., and perform semantic analysis on the texts, thereby assisting evaluators in quality evaluation. With the help of new-generation information technology means such as big data and artificial intelligence, the big data index method is used to propose an evaluation method for the quality of government texts, which reflects the effectiveness of texts in a more comprehensive, objective and intelligent way, and has an important impact on improving the quality of government texts and comprehensive information analysis capabilities.

[0055] The present invention is a method that does not rely entirely on manual labor. Figure 1-10As shown in the figure, a big data index for evaluating the quality of government texts is constructed by using big data and artificial intelligence technology. The index consists of a "1+n" index system, including 1 total index and n sub-indexes. The big data index is calculated by judging the degree of fit between government texts and various aspects at the semantic level, so as to reflect the response of government texts to topics such as national strategies, important policies, social conditions and public opinion, netizens' messages, and work plans. And it can give quantitative evaluation results for the degree of relevance. It is helpful to quantitatively reflect and evaluate the degree of implementation of government texts in implementing central policies, connecting key tasks, and reflecting social conditions and public opinion. And it can give objective and intelligent quality evaluation results for government texts from the perspective of relevance through intelligent algorithms, and can give quantitative evaluation results for the degree of relevance to evaluate the quality of government texts.

[0056] The present invention provides a method for evaluating the quality of government documents based on semantic analysis and big data index. The big data index is calculated by judging the degree of fit between the object to be evaluated and the benchmark reference at the semantic level. The index is composed of a "1+n" index system to reflect the response of the object to be evaluated to multiple topics such as national strategies, important policies, social conditions and public opinion, netizens' comments, work plans, etc. It includes the following steps:

[0057] Step 1: Set the benchmark data collection source and collect data according to the n dimensions that the government documents to be evaluated need to reflect;

[0058] Step 2: Text data preprocessing: performing text data preprocessing on the n benchmark text data sets collected in the first step, wherein the text data preprocessing includes text data denoising and deduplication;

[0059] Step 3: Natural language processing of benchmarking government text datasets and government text datasets to be evaluated;

[0060] The n benchmark text datasets obtained after the second step of text data preprocessing and the government text dataset to be evaluated are processed using natural language processing technology, including word segmentation, part-of-speech tagging, text vectorization, named entity recognition, keyword extraction, and text classification and clustering operations, to obtain n benchmark government text datasets corresponding to n sub-indexes;

[0061] Step 4: Calculation of n sub-indices;

[0062] Calculate the cosine similarity between the government text dataset to be evaluated and the n benchmark government text datasets in turn, and obtain the sub-index ρ in turn 分 ;

[0063] The cosine similarity is used to obtain the relevance between the government document to be evaluated and the n topics, which is the score index ρ 分 , the cosine similarity algorithm is as follows:

[0064]

[0065] A represents the government text vector to be evaluated, and B represents the benchmark subject vector. The closer the cosine value cosθ is to 1, the smaller the distance between the government text vector to be evaluated and the benchmark subject vector is, and the higher the degree of correlation is. Vectors A and B represent word frequency vectors. i , B i Represents the components of vectors A and B respectively, i is the dimension of vectors A and B, cosθ is the cosine similarity, and the index is ρ 分 ; The big data index is composed of a "1+n" index system, including a total index ρ 总 and n sub-indices ρ 分 ; n is a positive integer;

[0066] ρ 分 = cosθ

[0067] Step 5: Weight the n sub-indices to obtain the total index ρ 总 .

[0068] The advantages of the government text quality evaluation method based on semantic analysis and big data index provided by the present invention are:

[0069] 1. It can reflect the comprehensive change degree of complex factors that cannot be directly added. Through the "total index-sub-index" method, different indexes can be calculated based on different purposes to describe different aspects of the evaluation object, and can accurately locate a certain goal.

[0070] 2. It can evaluate the quality of texts at multiple scales and dimensions. First, it can measure the overall change trend of government documents of different sessions; second, it can measure the change and response degree of a given topic and describe the change trajectory of a given topic; third, it can compare between individual texts.

[0071] In summary, the government text quality evaluation method based on semantic analysis and big data index proposed in the present invention embodies the characteristics of data-driven decision-making in the digital age, and can reflect the comprehensive change degree of complex factors that cannot be directly added. Through the "total index-sub-index" method, different indexes can be calculated for different purposes, describing different aspects of the evaluation object, and can be accurately positioned for a certain target. Using the evaluation effect and feedback mechanism of the big data index, it can not only objectively reflect the overall change direction and degree of government texts (such as CPPCC proposals) between different sessions, but also reflect the direction and degree of influence of different topics on the overall change within a certain session, and even reflect the contribution of a single text to the evaluation index, which can be used as a beneficial supplement to the subjective evaluation method. Its advantages are manifested in the following three aspects: First, the traditional government text quality evaluation often relies on manual labor, and there are problems of subjectivity and difficulty in standardization. The implementation of the technical solution of the present invention can use intelligent means to perform automated evaluation, reduce the interference of human factors, and improve the objectivity of evaluation. Second, this is a standardized evaluation method that can ensure the uniformity and stability of the evaluation criteria and avoid evaluation bias caused by differences in personal understanding. Third, computers can be used to quickly process large amounts of government documents and improve evaluation efficiency. This advantage is particularly evident when processing massive amounts of data.

[0072] It should be noted that the method for obtaining the sub-index can be in this way, that is, comparing the spatial distances between the government text data set to be evaluated and n benchmark government text data sets in turn, and using cosine similarity calculation and normalization processing to obtain n sub-indexes in turn. In short, the big data index for evaluating the quality of government texts constructed by the present invention using big data and artificial intelligence technology can give objective and intelligent quality evaluation results for the quality of government texts from the perspective of relevance through intelligent algorithms.

[0073] Example 1: The value of n is 5.

[0074] In the first step, the benchmark data collection source is set to collect the five dimensions that need to be reflected in the government text to be evaluated, including: collecting target sample policies, target sample government work reports, target sample social public opinion information, target sample netizen message clues, and the subject content that needs to be reflected in the target sample topic information;

[0075] Step 2: Text data preprocessing: performing text data preprocessing on the benchmark text data set collected in the first step, wherein the text data preprocessing includes text data denoising and deduplication;

[0076] In the third step: natural language processing of the benchmark government text dataset and the government text dataset to be evaluated;

[0077] The n benchmark text data sets obtained after the second step of text data preprocessing and the government text data set to be evaluated are processed using natural language processing technology to obtain 5 benchmark government text data sets corresponding to the 5 sub-indices;

[0078] Step 4: Calculation of 5 sub-indices;

[0079] Define the index ρ 分 They are: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5; the total index ρ is obtained by weighting the five sub-indexes: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5. 总 .

[0080] Embodiment 2: The value of n is 2.

[0081] Take the policy responsiveness index ρ1 and the report fit index ρ2 as examples.

[0082] Big data index refers to all indices that use big data as the research object. It complements the traditional statistical survey index and is a relative numerical indicator used to reflect the comprehensive change degree of a phenomenon. The big data index combines the characteristics of big data and statistical index. It has similar mathematical, economic or sociological properties to traditional statistical survey indexes, and has big data properties that traditional statistical survey indexes do not have, reflecting extensiveness, continuity and refinement.

[0083] The government text quality evaluation method based on semantic analysis and big data index is characterized by comprising the following steps:

[0084] Step 1: Set the benchmarking data collection source, collect the two dimensions that need to be reflected in the government text to be evaluated, and obtain the benchmarking text data; specifically, collect the two dimensions that need to be reflected in the government text to be evaluated, including: collecting the target sample policies and the subject content that needs to be reflected in the target sample government work report;

[0085] Step 2: Text data preprocessing: performing text data preprocessing on the benchmark text data set collected in the first step, wherein the text data preprocessing includes text data denoising and deduplication;

[0086] Step 3: Natural language processing of benchmarking government text datasets and government text datasets to be evaluated;

[0087] The benchmark government text dataset and the government text dataset to be evaluated obtained after the second step of text data preprocessing are processed using natural language processing technology, including word segmentation, part-of-speech tagging, text vectorization, named entity recognition, keyword extraction, and text classification and clustering operations;

[0088] Step 4: Calculation of 2 sub-indices;

[0089] The cosine similarity between the government text dataset to be evaluated and the two benchmark government text datasets is calculated in turn, and the sub-index ρ is obtained in turn. 分 ;

[0090] The cosine similarity is used to obtain the relevance between the government document to be evaluated and the two topics, which is the score index ρ 分 , the cosine similarity algorithm is as follows:

[0091]

[0092] Step 5: Weight the two sub-indices: policy responsiveness sub-index ρ1 and report fit sub-index ρ2 to obtain the total index ρ 总 .

[0093] Policy responsiveness index ρ1:

[0094] Indicator acquisition: Comparison of the government text to be evaluated with national strategies, plans, important speeches of leaders, and national policies that can represent the central spirit at the semantic level.

[0095] Indicator explanation: This indicator is used to measure the response of the government documents to be evaluated to national strategies, plans, important speeches of leaders, and national policies. Data source: Obtain relevant national policies and regulations and the full text of leaders' speeches through authoritative websites.

[0096] Report fit index ρ2:

[0097] Indicator acquisition: Comparison of the government text to be evaluated with the government work reports, important documents, plans, important speeches of municipal leaders, etc. that can represent the key work of the province and city at the semantic level.

[0098] Indicator explanation: This indicator is used to measure the responsiveness of the proposal content to the key tasks of the province and city.

[0099] Data source: Obtain the full text of relevant provincial and municipal government policies, regulations, plans, and leaders’ speeches through authoritative websites.

[0100] "Public Concern" sub-index ρ3:

[0101] Indicator acquisition: Comparison of the government text to be evaluated with public opinion hotspots, social conditions and public opinion at the semantic level.

[0102] Indicator explanation: This indicator is used to measure the responsiveness of the proposal content to the basic situation of social life and the opinions and wishes of the people.

[0103] Data source: Information reflecting the trend of public opinion on the entire network, such as Weibo, WeChat, APP, Toutiao, web pages, forums, Q&A, newspapers, and TOP10 media reports related to the relevant proposals within a specified time period.

[0104] “Clue Utilization” sub-index ρ4:

[0105] Indicator acquisition: Comparison at the semantic level between the government text to be evaluated and the proposal clues collected and sorted based on social hot spots and netizens' comments.

[0106] Indicator explanation: This indicator is used to measure the response to message clues collected and sorted based on social hot spots and netizens' messages.

[0107] Data source: Message clues collected and sorted based on social hot spots and netizens’ messages.

[0108] "Topic participation" sub-index ρ5:

[0109] Indicator acquisition: Comparison at the semantic level between the government text to be evaluated and the topic suggestions given by relevant commissions and bureaus based on the government work plan.

[0110] Indicator explanation: This indicator is used to measure the response of the government documents to be evaluated to the topic suggestions given by relevant commissions and bureaus based on the government work plan.

[0111] Data source: Topic suggestions given by relevant commissions and bureaus based on the government work plan.

[0112] The proposal data used in the technical solution of the present invention can all come from the original proposal texts that can be made public in the proposal database of the Municipal CPPCC.

[0113] Among them, the n sub-indices are weighted to obtain the total index. The following weighting formula can be used, where i is a positive integer and the weight coefficient ɑ i The choice of and the number of sub-indices n are determined according to specific needs and can be adjusted according to needs.

[0114]

[0115] About index analysis: This paper takes the relevant conference data as an example to calculate and analyze the proposal big data index. There are 1532 proposals in the conference, with 97 keywords, such as Figure 5 As shown in (List of subject words for proposals in the government text quality evaluation method based on semantic analysis and big data index), for the convenience of statistical analysis, the subject words are semantically clustered to generate 11 categories. The corresponding relationship between the name and the number of proposals is as follows Figure 6 As shown, that is, Figure 6 The corresponding relationship between various names and the number of cases submitted in the government text quality evaluation method based on semantic analysis and big data index is clearly shown.

[0116] like Figure 7 As shown, after analyzing and sorting the 97 keywords, we got the 10 keywords with the highest total index, that is, the top 10 keywords are talent team, double reduction, digital economy, artificial intelligence, scientific and technological innovation, achievement transformation, epidemic prevention and control, medical emergency, business environment, and construction waste.

[0117] The ranking of the top 10 keywords in the government text quality evaluation method based on semantic analysis and big data index in the embodiment of the present invention in terms of policy responsiveness, report fit, public concern, clue utilization and topic participation is as follows: Figure 8 As shown in the table, the theme of "double reduction" has the highest response level to relevant policies and the highest level of consistency with relevant work reports. Figure 9-13 As shown, "talent team and scientific and technological innovation" have the highest public concern, and the public concern data for the talent team shows that it is as high as 0.98578178, indicating that the proposals on related topics have received a good response from the public opinion on the whole network; "scientific and technological innovation" has the highest level of clue utilization of the public's active suggestions and participation in the topic selection content with the organizer. Among all the keywords, the 10 themes with the highest policy responsiveness are, double reduction, caring for the disabled, childcare services, teacher team, intangible cultural heritage, fire management, renovation of old communities, ecological environment, civilized city, and property management. The policy responsiveness index of "double reduction" is as high as 1.000, "caring for the disabled" and "childcare services" are closely followed, with policy responsiveness indexes of 0.857 and 0.846 respectively. The policy responsiveness index of "property management" is the lowest, at 0.444.

[0118] Analysis by subject terms:

[0119] By analyzing the 11 subject categories, we can get the big data index ranked as scientific and technological innovation, talent services, epidemic prevention and control, urban construction, Beijing-Tianjin-Hebei coordination, others, elderly care services, transportation transformation, community services, intellectual property rights, and dual-carbon development. We can get the ranking of these 11 subject categories in terms of policy responsiveness from high to low, the ranking of these 11 subject categories in terms of report fit from high to low, the ranking of these 11 subject categories in terms of public concern from high to low, and the ranking of these 11 subject categories in terms of clue utilization from high to low.

[0120] like Fig.16The figure shows a schematic diagram of the policy responsiveness index of proposal subject classification in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention. Taking the policy responsiveness index and report fit index as examples for analysis, it can be seen that the community service subject has the best policy response, with a policy responsiveness index of 0.355, and the traffic transformation subject classification has the lowest policy responsiveness index, with the data showing 0.091.

[0121] like Fig.17 The figure shows a schematic diagram of the report fit index of proposal themes in the government text quality evaluation method based on semantic analysis and big data index in an embodiment of the present invention. It can be seen that the theme of "Beijing-Tianjin-Hebei coordination" has the best report fit, with the data showing 1.250, followed by "scientific and technological innovation" and "talent service", with the report fit index of "scientific and technological innovation" being 1.175, "talent service" being 1.173, and "dual carbon development" being the lowest, with the data showing 0.467.

[0122] It should be noted that the value of n can be a positive integer and is not limited to 5. n=5 is only a specific embodiment of the present invention and is within the protection scope of the present invention.

[0123] The specific embodiment of the present invention can analyze the subject words and subject classifications respectively, and compare them from five dimensions: policy responsiveness, report fit, public concern, clue utilization, and topic participation, as well as the weighted total index, to screen out "five good" subject words with good response to each dimension, and also screen out "single champion" subject words with outstanding performance in a single aspect. The relevant analysis results can be used to assist the CPPCC's related work and leadership decision-making.

[0124] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A government text quality evaluation method based on semantic analysis and big data index, characterized in that: The steps include: Step 1: Set the benchmark data collection source and collect data according to the n dimensions that the government documents to be evaluated need to reflect; Step 2: Text data preprocessing: performing text data preprocessing on the n benchmark text data sets collected in the first step, wherein the text data preprocessing includes text data denoising and deduplication; Step 3: Natural language processing of benchmarking government text datasets and government text datasets to be evaluated; The n benchmark text datasets obtained after the second step of text data preprocessing and the government text dataset to be evaluated are processed using natural language processing technology, including word segmentation, part-of-speech tagging, text vectorization, named entity recognition, keyword extraction, and text classification and clustering operations; Step 4: Calculation of n sub-indices; Calculate the cosine similarity between the government text dataset to be evaluated and the n benchmark government text datasets in turn, and obtain the sub-index ρ in turn 分 ; The cosine similarity is used to obtain the relevance between the government document to be evaluated and the n topics, which is the score index ρ 分 , the cosine similarity algorithm is as follows: A represents the government text vector to be evaluated, and B represents the benchmark subject vector. The closer the cosine value cosθ is to 1, the smaller the distance between the government text vector to be evaluated and the benchmark subject vector is, and the higher the degree of correlation is. Vectors A and B represent word frequency vectors. i , B i Represents the components of vectors A and B respectively, i is the dimension of vectors A and B, cosθ is the cosine similarity, and the index is ρ 分 ; The big data index is composed of a "1+n" index system, including a total index ρ 总 and n sub-indices ρ 分 ; n is a positive integer; r 分 =cosθ Step 5: Weight the n sub-indices to obtain the total index ρ 总 .

2. The government text quality evaluation method based on semantic analysis and big data index according to claim 1 is characterized in that: In the first step, the benchmark data collection source is set to collect the content of n dimensions that need to be reflected in the government text to be evaluated, n = 5, including: collecting the target sample policy, target sample government work report, target sample social public opinion information, target sample netizen message clues and target sample topic information. In the third step: benchmarking government text data set construction; The benchmark text data set obtained after the second step of text data preprocessing is processed using natural language processing technology to obtain 5 benchmark government text data sets corresponding to 5 sub-indices; define the sub-index ρ 分 They are: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5; the total index ρ is obtained by weighting the five sub-indexes: policy responsiveness index ρ1, report fit index ρ2, public concern index ρ3, clue utilization index ρ4, topic participation index ρ5. 总 .

Images