A method for pushing science and technology policies based on multidimensional semantic computing

By using multidimensional semantic computing methods, the matching degree between enterprise information and policy texts is analyzed, policy types are automatically identified and weights are adjusted, which solves the problem that enterprises have difficulty obtaining highly applicable policies and achieves efficient and accurate policy delivery.

CN123070118APending Publication Date: 2026-11-10NINGBO RENCE TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611239143.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-16
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

The science and technology policy information released by local governments is difficult for relevant enterprises to obtain directly, and enterprises cannot accurately judge the policy information with high policy relevance.

Method used

Using a multidimensional semantic computing method, the matching degree between enterprise information text and policy text is analyzed through the policy setting module and analysis module. The policy type is automatically identified and adaptive weight correction is performed. Finally, policies with high matching degree are selected and pushed.

Benefits of technology

This enables the accurate delivery of highly adaptable science and technology policies to relevant enterprises, improving the efficiency and accuracy of policy information utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FT_1
    Figure FT_1
  • Figure SMS_1
    Figure SMS_1
  • Figure SMS_10
    Figure SMS_10
Patent Text Reader

Abstract

This invention provides a method for pushing science and technology policies based on multidimensional semantic computing. The method includes the following steps: S1: Setting the policy tags and push scope for this push through the policy setting module; S2: Analyzing the science and technology policy text corresponding to the policy tags and the enterprise information text corresponding to the push scope through the analysis module to obtain an initial score for enterprise policy matching; S3: Identifying the policy type through the analysis module, performing adaptive weight correction and obtaining a corrected score for enterprise policy matching, filtering out enterprises for this push based on a threshold, and pushing the policies accordingly. This method can obtain the corresponding policies and the degree of matching through the score, allowing policies with high suitability to be pushed to the corresponding enterprises, and enterprises can also obtain policies with high suitability and appropriateness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a push method, specifically a science and technology policy push method based on multidimensional semantic computing. Background Technology

[0002] Currently, local governments across China are continuously issuing numerous science and technology support policies. However, these policies are typically published on their respective government websites, meaning businesses cannot directly access them and must actively search for them on the websites. Furthermore, it's difficult to accurately determine which policies are most relevant to their specific needs. Summary of the Invention

[0003] The present invention aims to at least partially solve one of the technical problems in related technologies: to provide a method for pushing science and technology policies based on multidimensional semantic computing, which obtains and pushes a set of specific science and technology policies with high matching scores with enterprises through science and technology policy text and enterprise information.

[0004] Therefore, one objective of this invention is to propose a method for pushing science and technology policies based on multidimensional semantic computing, the method comprising the following steps: S1: Configure the policy tags and push scope for this push notification through the policy settings module; S2: The analysis module analyzes the science and technology policy text corresponding to the policy tags and the enterprise information text corresponding to the push scope to obtain the initial score of enterprise policy matching. S3: The analysis module identifies policy types, performs adaptive weight adjustments, and obtains enterprise policy matching adjustment scores. Based on the threshold, it selects the enterprises to be pushed to and pushes them to the relevant departments.

[0005] Preferably, step S2 includes: preprocessing the science and technology policy text and the enterprise information text respectively, and extracting the policy text of each dimension through a pre-trained model; obtaining the initial matching score of the policy text of each dimension corresponding to the enterprise through a pre-trained semantic model, and normalizing the initial matching score of the policy text of each dimension to obtain the initial matching score of the enterprise policy.

[0006] Preferably, step S3 includes: automatically identifying policy types and loading initial weights through a semantic classification model; calculating the correction coefficients for each policy dimension to obtain temporary weights and final dynamic weights for each policy dimension; calculating the scores for each policy dimension and the enterprise policy matching correction score based on the final dynamic weights; and pushing the corresponding policy and score to the corresponding enterprise when the enterprise policy matching correction score is greater than the threshold.

[0007] Preferably, the calculation of the policy correction coefficients for each policy dimension to obtain the temporary weights and final dynamic weights for each policy dimension is as follows: calculate the constraint strength coefficients for each policy dimension, and combine the initial weights with the constraint strength coefficients for correction and normalization to obtain the final dynamic weights.

[0008] Preferably, the initial scores for text matching across all dimensions of the enterprise's corresponding policy are obtained using the following formula: Calculate the initial score S for text matching in the i-th dimension. i : Where i represents the policy dimension number; QUOTE For policy dimension vector, QUOTE For enterprise-level vectors, QUOTE With QUOTE The policy text and the corresponding enterprise text were vector-encoded using a pre-trained semantic model.

[0009] Preferably, the final dynamic weight is obtained from the formula: The final dynamic weights QUOTE are calculated. ; Where i represents the policy dimension number; QUOTE Initial weights; QUOTE Temporary weights for various policy dimensions; QUOTE Adjustment coefficients for various dimensions of the policy.

[0010] Preferably, the enterprise policy matching correction score is obtained by the formula: The calculated enterprise policy matching correction score is QUOTE. ; Where i represents the policy dimension number; QUOTE Scores for each dimension of the policy.

[0011] The preferred adjustment coefficients for each policy dimension are obtained using the following formula: T i2 = 3×n i1 +2×n i2 +n i3 T i3 = 3×m i1 +2×m i2 +m i3 The policy correction coefficients for each dimension were calculated. ; Where i represents the policy dimension number; QUOTE QUOTE QUOTE To integrate the weighting coefficients and satisfy QUOTE QUOTE To calculate the text frequency feature coefficients; QUOTE To calculate the sentence intensity characteristic coefficient; QUOTE To calculate the quantization coefficient for the level of precision; n i The frequency of relevant text fragments in the i-th policy dimension; three types of sentence keywords are defined: mandatory constraints, neutral support, and lenient guidance, n i1 Let n be the number of occurrences of mandatory constraint sentences under the i-th policy dimension. i2 Let n be the number of occurrences of the sexual support sentence structure under the i-th policy dimension. i3 T represents the number of occurrences of the easing guiding sentence structure under the i-th policy dimension; i2 The original total score for the sentence structure of the i-th policy dimension; three levels of information refinement are defined: Level 1 refinement, Level 2 refinement, and Level 3 refinement, m i1 m is the number of entities at the first level of refinement. i2 m is the number of entities at the second level of refinement. i3 The number of entities at level 3 refinement; T i3 The original total score represents the level of detail in the i-th policy dimension.

[0012] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

[0013] The above technical solution has the following advantages or beneficial effects: First, the analysis module analyzes the science and technology policy text corresponding to the policy tag, and analyzes the enterprise information text corresponding to the push scope to obtain the initial score of enterprise policy matching. The analysis module identifies the policy type, performs adaptive weight correction, and obtains the enterprise policy matching correction score. Based on the threshold, the enterprise to be pushed is selected and pushed. It can obtain the corresponding policy, and the degree of matching can be obtained through the score. It can push the policy with high matching degree to the corresponding enterprise, and the enterprise can also obtain the policy with high matching degree and appropriateness. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0015] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0016] The following describes in detail, with reference to the accompanying drawings, a method for pushing science and technology policies based on multidimensional semantic computing according to an embodiment of the present invention.

[0017] This invention provides a method for pushing science and technology policies based on multidimensional semantic computing, such as... Figure 1 As shown, S1: Configure the policy tags and push scope for this push notification through the policy settings module; S2: The analysis module analyzes the science and technology policy text corresponding to the policy tags and the enterprise information text corresponding to the push scope to obtain the initial score of enterprise policy matching. S3: The analysis module identifies policy types, performs adaptive weight adjustments, and obtains enterprise policy matching adjustment scores. Based on the threshold, it selects the enterprises to be pushed to and pushes them to the relevant departments.

[0018] The analysis module analyzes the enterprise information text corresponding to the push scope to obtain the initial score of enterprise policy matching. The analysis module identifies the policy type, performs adaptive weight correction, and obtains the enterprise policy matching correction score. Based on the threshold, the enterprises to be pushed are selected and pushed. The corresponding policies can be obtained, and the degree of correspondence can be obtained through the score.

[0019] In one embodiment of the present invention, step S2 includes: preprocessing the science and technology policy text and the enterprise information text respectively, and extracting the policy text of each dimension through a pre-trained model; obtaining the initial matching score of the enterprise's corresponding policy text of each dimension through a pre-trained semantic model, and normalizing the initial matching score of the policy text of each dimension to obtain the initial matching score of the enterprise policy. The science and technology policy text and the enterprise information text are preprocessed by cleaning, denoising, and sentence segmentation respectively; and the five unified dimensions of information are extracted from the policy text and the enterprise text respectively through a pre-trained information extraction model: industry field D1, geographical restriction D2, enterprise qualification D3, business scale D4, and technology direction D5.

[0020] As one embodiment of the present invention, step S3 includes: automatically identifying policy types and loading initial weights through a semantic classification model; calculating the correction coefficients for each dimension of the policy to obtain temporary weights and final dynamic weights for each dimension of the policy; calculating the scores for each dimension of the policy and the enterprise policy matching correction score based on the final dynamic weights; and pushing the corresponding policy and score to the corresponding enterprise when the enterprise policy matching correction score is greater than the threshold.

[0021] As one embodiment of the present invention, the temporary weights and final dynamic weights of each policy dimension are obtained by calculating the correction coefficients of each policy dimension: the constraint strength coefficients of each policy dimension are calculated, and the initial weights are corrected and normalized by combining the constraint strength coefficients to obtain the final dynamic weights.

[0022] As one embodiment of the present invention, the initial scores for text matching of each dimension of the enterprise's corresponding policy are obtained by the formula: Calculate the initial score S for text matching in the i-th dimension. i : Where i represents the policy dimension number; QUOTE For policy dimension vector, QUOTE For enterprise-level vectors, QUOTE With QUOTE The policy text and the corresponding enterprise text were vector-encoded using a pre-trained semantic model.

[0023] As one embodiment of the present invention, the final dynamic weight is obtained by the formula: The final dynamic weights QUOTE are calculated. ; Where i represents the policy dimension number; QUOTE Initial weights; QUOTE Temporary weights for various policy dimensions; QUOTE Adjustment coefficients for various policy dimensions. Automatically identify policy types through a semantic classification model: industry-oriented, qualification-supporting, locally implemented, R&D-innovation-oriented, and universally beneficial.

[0024] As one embodiment of the present invention, the enterprise policy matching correction score is obtained by the formula: The calculated enterprise policy matching correction score is QUOTE. ; Where i represents the policy dimension number; QUOTE Scores for each dimension of the policy.

[0025] As one embodiment of the present invention, the correction coefficients for each dimension of the policy are obtained by the following formula: T i2 = 3×n i1 +2×n i2 +n i3 T i3 = 3×m i1 +2×m i2 +m i3 The policy correction coefficients for each dimension were calculated. ; Where i represents the policy dimension number; QUOTE QUOTE QUOTE To integrate the weighting coefficients and satisfy QUOTE QUOTE To calculate the text frequency feature coefficients; QUOTE To calculate the sentence intensity characteristic coefficient; QUOTE To calculate the quantization coefficient for the level of precision; n i The frequency of relevant text fragments in the i-th policy dimension; three types of sentence keywords are defined: mandatory constraints, neutral support, and lenient guidance, n i1 Let n be the number of occurrences of mandatory constraint sentences under the i-th policy dimension. i2 Let n be the number of occurrences of the sexual support sentence structure under the i-th policy dimension. i3 T represents the number of occurrences of the easing guiding sentence structure under the i-th policy dimension; i2 The original total score for the sentence structure of the i-th policy dimension; three levels of information refinement are defined: Level 1 refinement, Level 2 refinement, and Level 3 refinement, m i1 m is the number of entities at the first level of refinement. i2 m is the number of entities at the second level of refinement. i3 The number of entities at level 3 refinement; T i3 The original total score represents the level of detail in the i-th policy dimension.

[0026] Preferably, a fusion weighting coefficient QUOTE can be set. 2. QUOTE QUOTE This demonstrates that the sentence intensity feature coefficient and the precision quantification coefficient are more important than the text frequency feature coefficient.

[0027] When setting keywords for three types of sentence structures: mandatory constraints, neutral support, and relaxed guidance. Keywords for mandatory constraint sentences can be: limited, must, require, should, must not, restrict, only, access, essential; Neutral supportive sentences can include keywords such as: support, encourage, meet, eligible for application, satisfy, and include. Keywords for loosely structured introductory sentences can be: priority, best choice, tendency, appropriate, recommendation, and encouragement of priority. For each policy dimension, entities for that dimension are extracted from the policy text using an information extraction model, including: specific geographical scope, enterprise qualification level, registered capital, number of employees, revenue scale, years of establishment, and specific technical field. When setting a level of refinement: Level 1, Level 2, and Level 3. The first-level detailed description includes specific values, a clear scope, hard indicators, and precise limitations. For example: the registered location is limited to a certain city, the registered capital is not less than 5 million, the company has been established for more than 2 years, it has high-tech enterprise qualifications, and it is limited to a specific sub-technical field.

[0028] The level 2 detailed description only has category restrictions and no precise values. It only divides the scope into broad categories, such as: registered enterprises within the province, enterprises with relevant industry qualifications, enterprises with a certain R&D scale, and enterprises belonging to high-tech related industries.

[0029] The Level 3 Detailed Level only briefly mentions the dimension name without any limitations or indicators, and the description is broad and vague, such as: meeting relevant business requirements, possessing the corresponding industry conditions, and having suitable technical capabilities.

[0030] In the description of this specification, references to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example that is included in at least one embodiment or example of the present invention.

[0031] In this specification, the illustrative expressions of the terms used do not necessarily refer to the same embodiments or examples. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples, without contradiction.

[0032] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

[0033] For those skilled in the art, various changes and modifications will undoubtedly be apparent after reading the above description. Therefore, the appended claims should be construed as covering all changes and modifications that encompass the true intent and scope of the invention. Any and all equivalent scope and content within the scope of the claims should be considered to remain within the intent and scope of the invention.

Claims

1. A method for pushing science and technology policies based on multidimensional semantic computing, characterized in that, The push method includes the following steps: S1: Configure the policy tags and push scope for this push notification through the policy settings module; S2: The analysis module analyzes the science and technology policy text corresponding to the policy tags and the enterprise information text corresponding to the push scope to obtain the initial score of enterprise policy matching. S3: The analysis module identifies policy types, performs adaptive weight adjustments, and obtains enterprise policy matching adjustment scores. Based on the threshold, it selects the enterprises to be pushed to and pushes them to the relevant departments.

2. The method for pushing science and technology policies based on multidimensional semantic computing according to claim 1, characterized in that, Step S2 includes: preprocessing the science and technology policy text and the enterprise information text respectively, and extracting the policy text of each dimension through a pre-trained model; obtaining the initial matching score of the policy text of each dimension corresponding to the enterprise through a pre-trained semantic model, and normalizing the initial matching score of the policy text of each dimension to obtain the initial matching score of the enterprise policy.

3. The method for pushing science and technology policies based on multidimensional semantic computing according to claim 2, characterized in that, Step S3 includes: automatically identifying policy types and loading initial weights through a semantic classification model; calculating the correction coefficients for each policy dimension to obtain temporary weights and final dynamic weights for each policy dimension; calculating the scores for each policy dimension and the enterprise policy matching correction score based on the final dynamic weights; and pushing the corresponding policy and score to the corresponding enterprise when the enterprise policy matching correction score is greater than the threshold.

4. The method for pushing science and technology policies based on multidimensional semantic computing according to claim 3, characterized in that, The temporary weights and final dynamic weights of each policy dimension are obtained by calculating the correction coefficients of each policy dimension: the constraint strength coefficients of each policy dimension are calculated, and the initial weights are corrected and normalized by combining the constraint strength coefficients to obtain the final dynamic weights.

5. A method for pushing science and technology policies based on multidimensional semantic computing according to claim 4, characterized in that, The initial scores for text matching across various dimensions of the enterprise's corresponding policies are obtained using the following formula: Calculate the initial score S for text matching in the i-th dimension. i : Where i is the policy dimension number; For policy dimension vector, For enterprise dimension vectors, and The policy text and the corresponding enterprise text were vector-encoded using a pre-trained semantic model.

6. The method for pushing science and technology policies based on multidimensional semantic computing according to claim 5, characterized in that, The final dynamic weights are obtained using the formula: The final dynamic weights are calculated. ; Where i is the policy dimension number; These are the initial weights; Temporary weights for each dimension of the policy; Adjustment coefficients for various dimensions of the policy.

7. A method for pushing science and technology policies based on multidimensional semantic computing according to claim 6, characterized in that, The adjusted score for enterprise policy matching is obtained using the following formula: Calculate the enterprise policy matching correction score ; Where i is the policy dimension number; Scores for each dimension of the policy.

8. A method for pushing science and technology policies based on multidimensional semantic computing according to claim 6, characterized in that, The adjustment coefficients for each dimension of the policy are obtained from the following formula: T i2 = 3×n i1 +2×n i2 +n i3 T i3 = 3×m i1 +2×m i2 +m i3 The policy correction coefficients for each dimension were calculated. ; Where i is the policy dimension number; , , To integrate the weighting coefficients, satisfy the following conditions: ; To calculate the text frequency feature coefficients; To calculate the sentence intensity characteristic coefficient; To calculate the quantization coefficient for the level of precision; n i The frequency of relevant text fragments in the i-th policy dimension; three types of sentence keywords are defined: mandatory constraints, neutral support, and lenient guidance, n i1 Let n be the number of occurrences of mandatory constraint sentences under the i-th policy dimension. i2 Let n be the number of occurrences of the sexual support sentence structure under the i-th policy dimension. i3 T represents the number of occurrences of the easing guiding sentence structure under the i-th policy dimension; i2 The original total score for the sentence structure of the i-th policy dimension; three levels of information refinement are defined: Level 1 refinement, Level 2 refinement, and Level 3 refinement, m i1 m is the number of entities at the first level of refinement. i2 m is the number of entities at the second level of refinement. i3 The number of entities at level 3 refinement; T i3 The original total score represents the level of detail in the i-th policy dimension.