Scientific and technological achievement management system for enterprise

Abstract

The invention relates to the technical field of scientific and technological achievement information management, and particularly discloses an enterprise scientific and technological achievement management system which specifically comprises the steps that a browsed text set and a non-browsed text set currently retrieved by a user are obtained in a calibration period based on a current retrieval scene, the text browsed by the user is divided into the target text and the non-target text through the behavior data, the target text and the non-browsed text are subjected to cosine similarity calculation, the feature value corresponding to the non-browsed text by the user is determined, the feature value is ranked and recommended to the user according to the interest feature value, and the reading efficiency of the user is improved by screening the text which the user is interested in. Therefore, the scientific and technological achievement management and conversion efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of scientific and technological achievement information management, and in particular to a scientific and technological achievement management system for an enterprise. Background Art

[0002] Scientific and technological achievements refer to intellectual products with certain recognized academic or economic value obtained through complex intellectual labor in scientific and technological activities. Scientific and technological achievements can achieve good economic, social or ecological environmental benefits. Their connotation is basically consistent with intellectual property rights and proprietary technologies, and they are an indispensable and important part of intangible assets.

[0003] At present, enterprises and research institutions generally use information technology to assist in the management of scientific and technological achievements, such as database systems, intellectual property management software, etc. Users can perform data mining and text analysis through keyword searches, and can filter out relevant texts from a large number of scientific and technological achievement documents.

[0004] With the rapid development of science and technology, the number of scientific and technological achievements has shown a significant growth trend. After searching by keywords, users are still faced with a large amount of text. Although these texts are highly similar in content, the key contents are different, which causes users to spend a lot of time comparing them one by one, thus affecting the management and transformation efficiency of scientific and technological achievements. Summary of the invention

[0005] The purpose of the present invention is to provide a management system for enterprise scientific and technological achievements to solve the above technical problems: The purpose of the present invention can be achieved through the following technical solutions: A system for managing scientific and technological achievements of an enterprise, characterized by comprising: The data acquisition module is used to acquire the browsed text set and the unbrowsed text set of the user's current search within the calibration period T, and acquire the number of interactive behaviors J and the viewing time T of the user on any browsed text, wherein the user interactive behaviors include clicks and slides; The text recommendation module is used to determine the interest value G corresponding to any browsed text based on the user's viewing time T and the number of user interactions J, and calculate the average viewing time The average number of interactions with users Calculate the threshold G', when G ≥ G', mark the text as the target text, when G < G', mark the text as the non-target text, generate a comprehensive text from all the target texts, preprocess the comprehensive text and any unbrowsed text respectively, obtain the preprocessed comprehensive text word vector set and the preprocessed any unbrowsed text word vector set, aggregate the comprehensive text word vector set and any unbrowsed text word vector set respectively, and obtain the comprehensive text overall vector A and the unbrowsed text overall vector D n , where n is any unbrowsed text; Get the overall vector A of the comprehensive text and the overall vector D of any unbrowsed text n Calculate the cosine value of the angle Q n , according to the cosine value of the angle Q n Obtain the characteristic values ​​X of all unbrowsed texts, sort all the unbrowsed texts from large to small according to the corresponding characteristic values ​​X, generate a sorting result, and recommend the unbrowsed texts to the user according to the sorting result.

[0006] As a further solution of the present invention: the data acquisition module also includes screening the text database according to the search terms input by the user to obtain search data, and sorting the search data according to the publication order of the texts and the relevance of the article titles to the search terms.

[0007] As a further solution of the present invention: in the data processing module, the preprocessing includes dividing the text into sentences, and performing word segmentation, stop word removal, and normalization on the divided text sentences.

[0008] As a further solution of the present invention: In the text recommendation module, the specific process of aggregating the word vector set is as follows: Embed all word vectors of different dimensions into the highest dimensional representation of the word vectors of all target texts and any word vectors of unbrowsed texts, according to the calculation formula , , where C a is any a-th word vector in all target texts, b is the total number of all target texts, C f is any f-th word vector in any unbrowsed text, and r is the total number of any unbrowsed text.

[0009] As a further solution of the present invention: In the text recommendation module, the specific process of obtaining the target text is as follows: The interest value G corresponding to any browsing text is determined by the user's viewing time T and the number of user interactions J. The interest value G corresponding to the browsing text is determined by the average viewing time The average number of interactions with users Calculate the threshold G´ using the following formula: ; ; G=T i +J i / y; G´= + / y; Wherein, i is any browsing text, y is a preset coefficient, when G≥G´, the text is marked as the target text, and when G<G´, the text is marked as the non-target text.

[0010] As a further solution of the present invention: in the text recommendation module, the calculation formula of the text cosine similarity is as follows: ; Among them, A*D n Represents vectors A and D n The dot product of |A| and |D n | represents vector A and vector D n The mold length.

[0011] As a further solution of the present invention: in the text recommendation module, the angle cosine value Q n According to the calculation formula X n =k*Q n Get the eigenvalue X n , where k is a preset coefficient and k>0, the eigenvalue X n Represents the feature value of the user for any n-th unbrowsed text.

[0012] As a further solution of the present invention: in the data acquisition module, if the start time of the calibration period is not the first search, all historical comprehensive text overall vectors in the historical calibration period are obtained, the cosine similarity Q' is calculated between the current comprehensive text overall vector A and all historical comprehensive text overall vectors, the historical comprehensive text overall vectors with Q'≥0.7 are screened out and marked as similar vectors, all historical comprehensive text word vector sets and the current comprehensive text word vector set corresponding to the similar text vectors are obtained, the two word vector sets are compared to screen out duplicate word vectors and generate a duplicate word vector set, the duplicate word vector set is aggregated to obtain an overall correction vector, and the overall correction vector is compared with any unbrowsed text overall vector D n The angle cosine value Q1 is calculated and the characteristic value X n , calculate the new eigenvalue X xn And sort all the unbrowsed texts from large to small according to the new feature value X, generate the sorting result, and recommend the corresponding text to the user according to the sorting result. The calculation formula is: ; Wherein, p and m are preset coefficients, and p>m>0; If the start time of the calibration period is the first retrieval, no correction is performed.

[0013] The beneficial effects of the present invention are as follows: based on the current search scenario, a set of browsed texts and a set of unbrowsed texts after the user's current search are obtained within a calibration period, user browsed texts are divided into target texts and non-target texts through behavioral data, cosine similarity is calculated between the target texts and the unbrowsed texts, the feature values ​​corresponding to the unbrowsed texts are determined, and the texts are recommended to the user according to the feature value sorting; if the start time of the calibration period is not the first search, a correction module is started to correct the current feature values, calculate new feature values, and recommend the unbrowsed texts to the user according to the feature value sorting; the reading efficiency of the user is improved by screening the texts that the user is interested in, thereby improving the efficiency of scientific and technological achievement management and transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings.

[0015] Figure 1 It is a structural schematic diagram of a system for managing scientific and technological achievements of an enterprise according to the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] The present invention is based on the scenario after the user searches through keywords, and performs secondary screening on the remaining unbrowsed texts through the user's reading of a large amount of texts obtained after the search. Based on the current search scenario, the browsed text set and the unbrowsed text set after the user's current search are obtained within a calibration period, and the user's browsed texts are divided into target texts and non-target texts through behavior data. The cosine similarity calculation is performed between the target text and the unbrowsed text to determine the feature value corresponding to the unbrowsed text by the user, and the texts are recommended to the user according to the feature value sorting. If the start time of the calibration period is not the first search, the correction module is turned on to correct the current feature value, calculate the new feature value, and recommend the unbrowsed texts to the user according to the feature value sorting. The reading efficiency of the user is improved by screening the texts that the user is interested in, thereby improving the efficiency of scientific and technological achievement management and transformation.

[0018] See also Figure 1 As shown, the present invention is a management system for enterprise scientific and technological achievements, characterized by comprising: The data acquisition module is used to acquire the browsed text set and the unbrowsed text set of the user's current search within the calibration period T, and acquire the number of interactive behaviors J and the viewing time T of the user on any browsed text, wherein the user interactive behaviors include clicks and slides; The text recommendation module is used to determine the interest value G corresponding to any browsed text based on the user's viewing time T and the number of user interactions J, and calculate the average viewing time The average number of interactions with users Calculate the threshold G', when G ≥ G', mark the text as the target text, when G < G', mark the text as the non-target text, generate a comprehensive text from all the target texts, preprocess the comprehensive text and any unbrowsed text respectively, obtain the preprocessed comprehensive text word vector set and the preprocessed any unbrowsed text word vector set, aggregate the comprehensive text word vector set and any unbrowsed text word vector set respectively, and obtain the comprehensive text overall vector A and the unbrowsed text overall vector D n , where n is any unbrowsed text; Get the overall vector A of the comprehensive text and the overall vector D of any unbrowsed text n Calculate the cosine value of the angle Q n , according to the cosine value of the angle Q n Obtain the characteristic values ​​X of all unbrowsed texts, sort all the unbrowsed texts from large to small according to the corresponding characteristic values ​​X, generate a sorting result, and recommend the unbrowsed texts to the user according to the sorting result.

[0019] It can be understood that the logic of judging whether a user is interested in a text by the number of user interaction behaviors J and the dwell time T is: when a user shows interest in a certain text segment, the interaction behaviors and dwell time with the text will increase.

[0020] It can be understood that the texts that the user is interested in are summarized into a comprehensive text, and its overall vector A is used as the central vector; then, the cosine similarity between the remaining unbrowsed text and the comprehensive text is calculated to obtain similar texts that the user is interested in; when it is not the first retrieval in the current calibration period, the historical comprehensive text vectors are obtained, and the overall vector A of the current comprehensive text is used as the central vector to screen out highly similar comprehensive texts; then, through the word vectors of all similar comprehensive texts and the word vector of the current comprehensive text, the repeated word vectors are screened out, and these repeated word vectors are aggregated into a revised overall vector; finally, with the revised overall vector as the central vector, the cosine similarity between the remaining unbrowsed text and it is recalculated, and the feature value of the current user is corrected.

[0021] In a preferred embodiment of the present invention, the data acquisition module further includes screening the text database according to the search terms input by the user to obtain search data, and sorting the search data according to the publication order of the texts and the relevance of the article titles to the search terms.

[0022] It can be understood that the present invention is based on the scenario after the user searches through keywords, and the text database is screened according to the search terms input by the user, which can ensure that the search data highly relevant to the user's needs is obtained, thereby improving the relevance and accuracy of the information.

[0023] In a preferred embodiment of the present invention, in the data processing module, the preprocessing includes dividing the text into sentences, and performing word segmentation, stop word removal, and normalization on the divided text sentences.

[0024] It is understandable that by dividing the text into sentences, the text structure can be analyzed and understood more accurately, thereby improving the accuracy of subsequent processing. Word segmentation and stop word removal can effectively extract key information from the text, reduce the interference of irrelevant content, and make subsequent data analysis more efficient. Normalization can unify different forms of vocabulary into a standard format, reduce the complexity brought by vocabulary diversity, and facilitate subsequent comparison and analysis.

[0025] In a preferred embodiment of the present invention, in the text recommendation module, the specific process of aggregating the word vector set is: Embed all word vectors of different dimensions into the highest dimensional representation of the word vectors of all target texts and any word vectors of unbrowsed texts, according to the calculation formula , , where C a is any a-th word vector in all target texts, b is the total number of all target texts, C f is any f-th word vector in any unbrowsed text, and r is the total number of any unbrowsed text.

[0026] It can be understood that embedding word vectors of different dimensions into a unified high-dimensional representation ensures data consistency and facilitates subsequent calculations and comparisons. The aggregation process can reduce computational complexity, allowing the system to generate recommendation results more quickly when processing large-scale text.

[0027] In a preferred embodiment of the present invention, in the text recommendation module, the specific process of obtaining the target text is as follows: The interest value G corresponding to any browsing text is determined by the user's viewing time T and the number of user interactions J. The interest value G corresponding to the browsing text is determined by the average viewing time The average number of interactions with users Calculate the threshold G´ using the following formula: ; ; G=T i +J i / y; G´= + / y; Wherein, i is any browsing text, y is a preset coefficient, when G≥G´, the text is marked as the target text, and when G<G´, the text is marked as the non-target text.

[0028] It is understandable that calculating the interest value by viewing time and number of interactions can more accurately reflect the user's real interest in the text and enhance the personalization of recommendations. Using the mean to calculate the dynamic threshold G' can adapt to the behavior patterns and preferences of different users, making the recommendation system more flexible and adaptable. By focusing on the content that users are really interested in, the interference of irrelevant information can be reduced, and the user's satisfaction and usage experience can be improved.

[0029] In a preferred embodiment of the present invention, in the text recommendation module, the calculation formula of text cosine similarity is as follows: ; Among them, A*D n Represents vectors A and D n The dot product of |A| and |D n | represents vector A and vector D n The mold length.

[0030] In a preferred embodiment of the present invention, in the text recommendation module, the angle cosine value Q n According to the calculation formula X n =k*Q n Get the eigenvalue X n , where k is a preset coefficient and k>0, the eigenvalue X n Represents the feature value of the user for any n-th unbrowsed text.

[0031] In a preferred embodiment of the present invention, in the data acquisition module, if the start time of the calibration period is not the first search, all historical comprehensive text overall vectors in the historical calibration period are obtained, the cosine similarity Q' is calculated between the current comprehensive text overall vector A and all historical comprehensive text overall vectors, the historical comprehensive text overall vectors with Q'≥0.7 are screened out and marked as similar vectors, all historical comprehensive text word vector sets and the current comprehensive text word vector set corresponding to the similar text vectors are obtained, the two word vector sets are compared to screen out duplicate word vectors and generate a duplicate word vector set, the duplicate word vector set is aggregated to obtain an overall correction vector, and the overall correction vector is compared with any unbrowsed text overall vector D n The angle cosine value Q1 is calculated and the characteristic value X n , calculate the new eigenvalue X xn And sort all the unbrowsed texts from large to small according to the new feature value X, generate the sorting result, and recommend the corresponding text to the user according to the sorting result. The calculation formula is: ; Wherein, p and m are preset coefficients, and p>m>0; If the start time of the calibration period is the first retrieval, no correction is performed.

[0032] It is understandable that by obtaining historical comprehensive text vectors, existing data can be used to improve the accuracy of current recommendations and enhance the learning ability of the system. Using cosine similarity Q' to screen similar texts can effectively identify historical target texts that are semantically similar to the current target text, thereby improving the relevance of recommendations. Comparing and filtering duplicate word vectors and generating overall correction vectors can reduce redundant information, making the final recommendation results more refined and effective.

[0033] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A system for managing enterprise scientific and technological achievements, characterized in that: include: The data acquisition module is used to acquire the browsed text set and the unbrowsed text set of the user's current search within the calibration period T, and acquire the number of interactive behaviors J and the viewing time T of the user on any browsed text, wherein the user interactive behaviors include clicks and slides; The text recommendation module is used to determine the interest value G corresponding to any browsed text based on the user's viewing time T and the number of user interactions J, and calculate the average viewing time The average number of interactions with users Calculate the threshold G', when G ≥ G', mark the browsed text as the target text, when G < G', mark the browsed text as the non-target text, aggregate all the target texts into a comprehensive text, preprocess the comprehensive text and any unbrowsed text respectively, obtain the preprocessed comprehensive text word vector set and the preprocessed any unbrowsed text word vector set, aggregate the comprehensive text word vector set and any unbrowsed text word vector set respectively, and obtain the comprehensive text overall vector A and the unbrowsed text overall vector D n , where n is any unbrowsed text; Get the overall vector A of the comprehensive text and the overall vector D of any unbrowsed text n Calculate the cosine value of the angle Q n , according to the cosine value of the angle Q n Obtain the characteristic values ​​X of all unbrowsed texts, sort all the unbrowsed texts from large to small according to the corresponding characteristic values ​​X, generate a sorting result, and recommend the unbrowsed texts to the user according to the sorting result.

2. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: The data acquisition module also includes screening the text database according to the search terms input by the user to obtain search data, and sorting the search data according to the publication order of the texts and the relevance between the article titles and the search terms.

3. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the data processing module, the preprocessing includes dividing the text into sentences, and performing word segmentation, stop word removal, and normalization on the divided text sentences.

4. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the text recommendation module, the specific process of aggregating the word vector set is as follows: Embed all word vectors of different dimensions into the highest dimensional representation of the word vectors of all target texts and any word vectors of unbrowsed texts, according to the calculation formula , , where C a is any a-th word vector in all target texts, b is the total number of all target texts, C f is any f-th word vector in any unbrowsed text, and r is the total number of any unbrowsed text.

5. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the text recommendation module, the specific process of obtaining the target text is as follows: The interest value G corresponding to any browsing text is determined by the user's viewing time T and the number of user interactions J. The interest value G corresponding to the browsing text is determined by the average viewing time The average number of interactions with users Calculate the threshold G´ using the following formula: ； ； G=T i +J i / y; G´= + / y; Wherein, i is any browsing text, y is a preset coefficient, when G≥G´, the text is marked as the target text, and when G<G´, the text is marked as the non-target text.

6. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the text recommendation module, the calculation formula of text cosine similarity is as follows: ； Among them, A*D n Represents vectors A and D n The dot product of |A| and |D n | represents vector A and vector D n The mold length.

7. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the text recommendation module, the angle cosine value Q n According to the calculation formula X n =k*Q n Get the eigenvalue X n , where k is a preset coefficient and k>0, the eigenvalue X n Represents the feature value of the user for any n-th unbrowsed text.

8. The system for managing enterprise scientific and technological achievements according to claim 1, characterized in that: In the data acquisition module, if the start time of the calibration period is not the first search, all historical comprehensive text overall vectors in the historical calibration period are obtained, the cosine similarity Q' is calculated between the current comprehensive text overall vector A and all historical comprehensive text overall vectors, the historical comprehensive text overall vectors with Q'≥0.7 are screened out and marked as similar vectors, all historical comprehensive text word vector sets and the current comprehensive text word vector set corresponding to the similar text vectors are obtained, the two word vector sets are compared to screen out duplicate word vectors and generate a duplicate word vector set, the duplicate word vector set is aggregated to obtain an overall correction vector, and the overall correction vector is compared with any unbrowsed text overall vector D n The angle cosine value Q1 is calculated and the characteristic value X n , calculate the new eigenvalue X xn And sort all the unbrowsed texts from large to small according to the new feature value X, generate the sorting result, and recommend the corresponding text to the user according to the sorting result. The calculation formula is: ; Wherein, p and m are preset coefficients, and p>m>0; If the start time of the calibration period is the first retrieval, no correction is performed.

Images