Data push method and system

Through clustering and semantic analysis technology, combined with user identification classification methods, the problem of insufficient accuracy and comprehensiveness of scientific and technological service resource recommendations in the existing technology is solved, and more accurate and comprehensive recommendations are achieved, improving user experience and corporate competition are improved. force.

CN117874361BActive Publication Date: 2024-11-08SHENZHEN JISHENGCHANG IND CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410075255.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-18
Publication Date
2024-11-08
Estimated Expiration
2044-01-18

AI Technical Summary

Technical Problem

When recommending scientific and technological service resources, the existing technology lacks accuracy and comprehensiveness and cannot meet the diversified needs of corporate users.

Method used

By obtaining the search content input by the user, traversing the database to obtain question statements and reply statements with high similarity, performing clustering and semantic analysis, generating prompt information, guiding users to enter the complete search content, and dividing user sets according to the user's identity, recommending similar preferences technology service information.

Benefits of technology

It has achieved more accurate and comprehensive recommendation of scientific and technological service resources, improved user experience and corporate competitiveness, and met the actual and potential needs of users through a combination of passive and active recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117874361B_ABST
    Figure CN117874361B_ABST
Patent Text Reader

Abstract

The present invention discloses a data push method and system: traverse a first database based on a first search content input by a user, send a prompt message to the first user according to the matching result, receive a second search content input by the first user according to the prompt message, search and obtain first push data according to the second search content; obtain the first search records of all users in the user set where the first user is located within a first preset time, and use the browsing data in the first search record with a click rate higher than a second preset value as the second push data; filter and sort the push data according to preset rules and send it to the user terminal, wherein the push data includes the first push data and the second push data. The technical solution of the present invention realizes a service recommendation method that combines passive recommendation and active recommendation, and makes accurate and comprehensive service recommendations based on the actual and potential needs of users.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet search recommendation, and in particular relates to a data push method and system. Background Art

[0002] The science and technology service industry is a new industry classification formed under the trend of continuous refinement of division of labor and continuous integration and growth of industries today. It includes eight specialized services: research and development, inspection and testing certification, entrepreneurship incubation, intellectual property, science and technology consulting, science and technology finance, technology transfer, and popularization of science and technology.

[0003] With the development of Internet technology, corporate users have an increasing demand for obtaining information on scientific and technological services. However, when faced with a large number of service providers, service types, and service solutions, it is very important for enterprises to obtain scientific and technological services that are suitable for themselves from a large amount of information. In the prior art, for example, Chinese patent application CN113221001A discloses a method for intelligent matching and recommendation of scientific and technological services based on big data, which organizes and classifies the business content of the enterprise by the first keyword; retrieves according to the second keyword entered by the enterprise and displays it through the public display module; performs statistical calculations on the frequency of occurrence of the first keyword and the second keyword, and displays them separately on one side of the search engine module through the recommendation matching module, so that the enterprise can perform double screening according to its own needs and recommendations. This method recommends scientific and technological services based on the keywords entered by the user and the keywords of the business content, which is not comprehensive enough. Another example is Chinese patent application CN115374354A, which discloses a method, device, equipment and medium for recommending science and technology services based on machine learning, which obtains historical transaction records of science and technology services and the target needs of users, analyzes the target needs of users based on the attributes, behaviors and search engine retrieval rules of users, defines the characteristic labels of the target needs and historical transaction records of science and technology services, uses machine learning algorithms to perform machine training on the characteristic labels, determines science and technology service resources that match the target needs based on the training results, and recommends the science and technology service resources to users. This method recommends science and technology service resources based on historical transaction records and target needs of users, which is not comprehensive enough.

[0004] Therefore, it is an urgent problem to provide a data push method and system to recommend scientific and technological service resource data to users more accurately and comprehensively. Summary of the invention

[0005] In view of the above-mentioned technical problems, the present invention provides a data push method and system.

[0006] In a first aspect, the present invention provides a data push method, the method comprising:

[0007] Obtain a first search content input by a first user, traverse a first database based on the first search content, obtain a first question statement having a similarity with the first search content greater than a first preset value, further obtain N1 first answer statements corresponding to the first question statement in the first database, cluster the N1 first answer statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to a user terminal of the first user, receive a second search content input by the first user according to the prompt information, and then search according to the second search content to obtain first push data;

[0008] Obtain a first user identifier of a first user, determine whether the first user is in any user set based on the first user identifier, if the first user is in any user set, obtain first search records of all users in any user set within a first preset time, and use browsing data with a click rate higher than a second preset value in the first search record as second push data; if the first user is not in, obtain a first historical comment data set of the first user, calculate the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, divide the first user into corresponding user sets based on the comment similarity, then obtain second search records of all users in the corresponding user set within the first preset time, and use browsing data with a click rate higher than the second preset value in the second search record as second push data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is browsing data commented on by the first user, and the second user is a user other than the first user;

[0009] The push data is screened and sorted according to preset rules and sent to the user terminal, wherein the push data includes first push data and second push data.

[0010] Specifically, the steps of establishing the first database include:

[0011] Step 111: traverse the social platforms connected to the data recommendation terminal to obtain search information input by any user, and if the search information meets the first preset condition, define the search information as a question statement to be selected;

[0012] Step 112: Obtain all answer information corresponding to the question statement to be selected, and if any answer information meets the second preset condition, define any answer information as the first answer statement to be selected;

[0013] Step 113: Calculate the answer similarity between any two first answer statements to be selected. If the answer similarity is greater than a fourth preset value, delete any two first answer statements to be selected. After traversing all first answer statements to be selected, define the remaining first answer statements to be selected as answer statements to be selected.

[0014] Step 114: after traversing all social platforms, the question similarity between any two question statements to be selected is calculated. If the question similarity is greater than a fifth preset value, any one of the two question statements to be selected is deleted, and the answer statements to be selected corresponding to the two question statements to be selected are merged. After the merging is completed, the duplicate answer statements to be selected are deleted.

[0015] Step 115: After traversing all the question statements to be selected, define the remaining answer statements to be selected as question statements, and define the answer statements to be selected corresponding to the question statements as answer statements, and then store the question statements and the answer statements corresponding to the question statements in correspondence to generate a first database.

[0016] Specifically, the first preset condition is that the search information contains question words;

[0017] The second preset condition is that the similarity between any answer information and the sentence of the question to be selected is less than a sixth preset value and any answer information is a supplement to the sentence of the question to be selected.

[0018] Specifically, clustering the N1 first reply sentences and performing semantic analysis on the clustering results to generate prompt information includes the following steps:

[0019] Step 121, calculating the first similarity between the first search content and any first reply statement, if the first similarity is greater than a seventh preset value, deleting any first reply statement, and after traversing all first reply statements, defining the remaining first reply statements as second reply statements;

[0020] Step 122: cluster the second reply statement to generate N2 reply clusters;

[0021] Step 123: Use N3 decision trees to score any second reply statement, and use the average of the score values ​​of all decision trees as the statement score of any second reply statement;

[0022] Step 124: Calculate the cluster score of any reply cluster based on the sentence scores of all second reply sentences in any reply cluster. The calculation formula is:

[0023]

[0024] Among them, CS i is the clustering score of the i-th answer cluster, N4 is the total number of second answer statements, Mi is the number of second reply statements in the i-th reply cluster, R j is the sentence score of the jth second reply sentence in the i-th reply cluster;

[0025] Step 125: sort all reply clusters based on the cluster scores, perform semantic analysis on the top N5 reply clusters, generate N5 prompt messages and send them to the user terminal.

[0026] Specifically, obtaining a first historical review data set of a first user, calculating the review similarity between the first user and the second user based on the first historical review data set and the second historical review data set, and dividing the first user into a corresponding user set based on the review similarity includes:

[0027] Step 21, obtaining browsing data corresponding to any review data in the first historical review data set;

[0028] Step 22: determine whether the browsing data corresponding to any review data is a discussion on a specific topic or a free discussion. If it is a discussion on a specific topic, extract the keywords of the specific topic as the first category keywords of any review data. If it is a free discussion, extract the first category keywords corresponding to any review data from any review data, wherein the first category keywords are words related to scientific and technological services.

[0029] Step 23: perform semantic analysis on any review data to determine whether any review data is a positive review or a negative review. If it is a negative review, store the first category keywords corresponding to any review data in the first storage module. If it is a positive review, extract all the second category keywords in any review data and calculate the first review value of any review data. The calculation formula is:

[0030]

[0031] Among them, S1 is the first review value of any review data, W k is the weight of the kth second-category keyword, N6 is the number of sentences in any review data, and N k is the ranking of the sentences where the kth second-category keyword is located, K1 is the total number of second-category keywords in any review data, and the second-category keywords are positive review words;

[0032] Step 24, traversing all the review data in the first historical review data set, when there are identical first-category keywords, taking the average of the first review values ​​corresponding to the identical first-category keywords as the first final review value of the identical first-category keywords, and then correspondingly storing the first user identifier, any first-category keyword, and the first final review value corresponding to any first-category keyword in the second storage module;

[0033] Step 25: extract the second user identifier of any second user in the second historical review data set, obtain browsing data corresponding to any review data of any second user in the second historical review data set, and calculate the second final review value corresponding to any first-category keyword by any second user according to the method of steps 22 to 24;

[0034] Step 26: After traversing all second users in the second historical review data set, any second user identifier, any first-category keyword, and a second final review value corresponding to any first-category keyword are stored in the second storage module in a corresponding manner;

[0035] Step 27: define the first category of keywords corresponding to the first final review value greater than the eighth preset value as keywords to be recommended, and calculate the similarity of the reviews of the first user and any second user on any keyword to be recommended based on the first final review value of any keyword to be recommended by the first user and the second final review value of any keyword to be recommended by any second user; when the review similarity is greater than the ninth preset value, classify the first user and any second user into the same user set.

[0036] Specifically, before acquiring the first historical review data set and the second historical review data set, a unique authentication identifier is allocated to any user according to the real-name authentication information of any user, and the historical review data of any user is collected based on the authentication identifier.

[0037] Specifically, filtering and sorting the push data according to the preset rules and sending the push data to the user terminal includes:

[0038] Step 31, determining whether the first category of keywords in any push data title exists in the first storage module, if so, deleting any push data, and after traversing all push data, defining the remaining push data as data to be pushed;

[0039] Step 32: define the first category of keywords in any title of the data to be pushed as keywords to be sorted, and determine whether the gth keyword to be sorted exists in the second storage module. If so, proceed to step 33; if not, proceed to step 34, where g is a positive integer from 1 to G, and G is the total number of keywords to be sorted;

[0040] Step 33: Based on the second final comment value of any second user on the g-th keyword to be sorted, and the similarity of comments between any second user and the first user on the g-th keyword to be sorted, the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended is calculated, and the calculation formula is:

[0041]

[0042] Among them, RL gis the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended, SC gq C2 is the similarity between the comments of the qth second user and the first user on the gth keyword to be recommended, gq is the second final comment value of the qth second user on the gth keyword to be recommended, and Q is the total number of second users who comment on the gth keyword to be recommended;

[0043] Step 34: Obtain the click rate of the data to be pushed corresponding to the gth keyword to be sorted;

[0044] Step 35: After traversing all the keywords to be sorted, sort the data to be pushed based on the recommendation degree and / or click rate of the data to be pushed.

[0045] In a second aspect, the present invention further provides a data push system, the system comprising:

[0046] A data push terminal and a user terminal, wherein the data push terminal includes a data receiving module, a first push module, a second push module, and a data output module;

[0047] A data receiving module, used to obtain a first search content input by a first user;

[0048] A first push module is used to traverse the first database according to the first search content, obtain a first question statement whose similarity with the first search content is greater than a first preset value, further obtain N1 first answer statements corresponding to the first question statement in the first database, cluster the N1 first answer statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to the user terminal of the first user, receive the second search content input by the first user according to the prompt information, and then search according to the second search content to obtain the first push data;

[0049] A second push module is used to obtain a first user identifier of a first user, determine whether the first user is in any user set according to the first user identifier, and if so, obtain first search records of all users in any user set within a first preset time, and use browsing data with a click rate higher than a second preset value in the first search record as second push data; if not, obtain a first historical comment data set of the first user, calculate the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, and divide the first user into corresponding user sets based on the comment similarity, and then obtain second search records of all users in the corresponding user set within the first preset time, and use browsing data with a click rate higher than the second preset value in the second search record as second push data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is browsing data commented on by the first user, and the second user is a user other than the first user;

[0050] The data output module is used to filter and sort the push data according to preset rules and send the push data to the user terminal, wherein the push data includes first push data and second push data.

[0051] The present invention discloses a data push method and system. The method comprises the following steps: traversing a first database based on first search content input by a first user, obtaining a first question statement whose similarity with the first search content is greater than a first preset value, further obtaining N1 first answer statements corresponding to the first question statement in the first database, clustering the N1 first answer statements, performing semantic analysis on the clustering result to generate prompt information, and sending the prompt information to a user terminal of the first user; receiving a second search content input by the first user according to the prompt information, guiding the first user to input the complete search content, improving the search precision and accuracy, and then searching and obtaining first push data according to the re-entered search content; obtaining a first user identifier of the first user, judging whether the first user is in any user set according to the first user identifier, and if not, obtaining a first historical comment data set of the first user, calculating the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, dividing the first user into a corresponding user set based on the comment similarity, and if the first user is in any user set within a first preset time, obtaining first search records of all users in any user set, and using browsing data in the first search record with a click rate higher than a second preset value as second push data, and based on this, recommending to the first user science and technology service information concerned by other users with similar preferences.

[0052] The technical solution of the present invention realizes a service recommendation method that combines passive recommendation and active recommendation, and provides enterprises with better service experience and enhances their competitiveness through precise matching and personalized recommendations based on users' actual needs and potential needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0054] Figure 1 A flow chart of a data push method of the present invention;

[0055] Figure 2 It is a structural schematic diagram of a data push system of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the specific embodiments described herein are only used to explain the present invention and are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0057] It should be noted that if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0058] Figure 1 FIG. 1 is a flow chart of an embodiment of a data push method provided by the present invention, and the flow chart specifically includes:

[0059] Obtain a first search content input by a first user, traverse the first database based on the first search content, obtain a first question statement whose similarity with the first search content is greater than a first preset value, further obtain N1 first reply statements corresponding to the first question statement in the first database, cluster the N1 first reply statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to the user terminal of the first user, receive a second search content input by the first user according to the prompt information, and then search according to the second search content to obtain first push data.

[0060] The first preset value is set according to the experience of those skilled in the art or according to an actual application scenario, and the embodiments of the present application are not limited to this.

[0061] Exemplarily, the similarity between the first search content and the question sentence is calculated based on the sentence vectors of the first search content and the question sentence.

[0062] The first database stores question statements and N1 answer statements corresponding to the question statements, assigns identification information to each question statement, and obtains the answer statement corresponding to the question statement from the first database based on the identification information of the question statement, wherein N1 is a positive integer greater than or equal to 1.

[0063] Exemplarily, the clustering algorithm is a K-means algorithm, and the semantic analysis model is a NLP algorithm.

[0064] When users search for information through search platforms, the search content they input covers a wide range, resulting in a large number of search results. To obtain the target data they want, they need to browse a large amount of data, which is time-consuming and laborious. Therefore, before searching, guide users to enter the complete search content.

[0065] Specifically, the steps of establishing the first database include:

[0066] Step 111: traverse the social platforms connected to the data recommendation terminal to obtain search information input by any user. If the search information meets the first preset condition, the search information is defined as a question statement to be selected.

[0067] Step 112: Obtain all answer information corresponding to the question statement to be selected. If any answer information meets the second preset condition, define any answer information as the first answer statement to be selected.

[0068] Step 113, calculate the response similarity between any two first candidate response statements. If the response similarity is greater than a fourth preset value, delete any two first candidate response statements. After traversing all first candidate response statements, define the remaining first candidate response statements as candidate response statements.

[0069] Step 114: After traversing all social platforms, calculate the question similarity between any two candidate question statements. If the question similarity is greater than the fifth preset value, delete any two candidate question statements, and merge the candidate answer statements corresponding to any two candidate question statements. After the merging is completed, delete the duplicate candidate answer statements.

[0070] Step 115: After traversing all the question statements to be selected, define the remaining answer statements to be selected as question statements, and define the answer statements to be selected corresponding to the question statements as answer statements, and then store the question statements and the answer statements corresponding to the question statements in correspondence to generate a first database.

[0071] Specifically, the first preset condition is that the search information contains question words;

[0072] The second preset condition is that the similarity between any answer information and the sentence of the question to be selected is less than a sixth preset value and any answer information is a supplement to the sentence of the question to be selected.

[0073] The first database is established by traversing the social platforms connected to the data recommendation terminal. The generated first database is more professional and has a strong relevance to the target data to be searched.

[0074] Preferably, if the question similarity between any two candidate question statements is greater than a fifth preset value, any one of the two candidate question statements is deleted, and after the candidate answer statements corresponding to any two candidate question statements are merged, the duplicate candidate answer statements are deleted based on the similarity between any two candidate answer statements.

[0075] Preferably, semantic analysis is performed to determine whether any answer information is related to a candidate question statement corresponding to any of the answer information, and whether it contains a supplementary statement of the candidate question statement.

[0076] Preferably, the second preset condition also includes whether any answer information is related to any of the candidate question statements corresponding to the answer information.

[0077] For example, in the first database, the question statement is "What scientific and technological services are available?", and the corresponding answer statements are "There are many scientific and technological services, specifically involving eight areas", "Different scientific and technological service providers provide different scientific and technological services", etc.

[0078] Specifically, clustering the N1 first reply sentences and performing semantic analysis on the clustering results to generate prompt information includes the following steps:

[0079] Step 121, calculate the first similarity between the first search content and any first reply statement. If the first similarity is greater than the seventh preset value, delete any first reply statement. After traversing all first reply statements, define the remaining first reply statements as second reply statements.

[0080] Step 122: cluster the second reply statement to generate N2 reply clusters.

[0081] Step 123: Use N3 decision trees to score any second reply statement, and take the average of the score values ​​of all decision trees as the statement score of any second reply statement.

[0082] Step 124: Calculate the cluster score of any reply cluster based on the sentence scores of all second reply sentences in any reply cluster. The calculation formula is:

[0083]

[0084] Among them, CS i is the clustering score of the i-th answer cluster, N4 is the total number of second answer statements, M i is the number of second reply statements in the i-th reply cluster, R j is the sentence score of the jth second reply sentence in the i-th reply cluster.

[0085] Step 125: sort all reply clusters based on the cluster scores, perform semantic analysis on the top N5 reply clusters, generate N5 prompt messages and send them to the user terminal.

[0086] Through decision tree voting, any reply statement is quantified to obtain the sentence score, and then the clustering score of any reply cluster is calculated based on the sentence score. The semantic analysis of the top N5 reply clusters with high clustering scores is performed to generate prompt information.

[0087] For example, the first search content input by the first user is "technology services", and the corresponding prompt information is "which field of technology services do you want", "which provider of technology services", etc. After receiving the above prompt information, the first user enters the second search content according to his own needs, such as "technology services in the field of technology consulting".

[0088] According to the technical solution of the present invention, prompt information can be generated for the user, so that the user can input more complete and accurate search content, thereby improving the search accuracy and efficiency.

[0089] A first user identifier of a first user is obtained, and whether the first user is located in any user set is determined based on the first user identifier. If the first user is located in any user set, first search records of all users in any user set within a first preset time are obtained, and browsing data with a click rate higher than a second preset value in the first search record is used as second push data; if the first user is not located, a first historical comment data set of the first user is obtained, and the comment similarity between the first user and the second user is calculated based on the first historical comment data set and the second historical comment data set, and the first user is divided into corresponding user sets based on the comment similarity, and then, second search records of all users in the corresponding user set within the first preset time are obtained, and browsing data with a click rate higher than the second preset value in the second search record is used as second push data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is browsing data commented on by the first user, and the second user is a user other than the first user.

[0090] Exemplarily, when the user inputs the search content, or when the user logs into the search platform, the retrieval of the second pushed data for the user begins.

[0091] Preferably, the user set is updated periodically for each user.

[0092] Specifically, obtaining a first historical review data set of a first user, calculating the review similarity between the first user and the second user based on the first historical review data set and the second historical review data set, and dividing the first user into a corresponding user set based on the review similarity includes:

[0093] Step 21: Obtain browsing data corresponding to any review data in the first historical review data set.

[0094] Step 22, determine whether the browsing data corresponding to any review data is a discussion on a specific topic or a free discussion. If it is a discussion on a specific topic, extract the keywords of the specific topic as the first category keywords of any review data. If it is a free discussion, extract the first category keywords corresponding to any review data from any review data, wherein the first category keywords are words related to technology services.

[0095] Step 23: perform semantic analysis on any review data to determine whether any review data is a positive review or a negative review. If it is a negative review, store the first category keywords corresponding to any review data in the first storage module. If it is a positive review, extract all the second category keywords in any review data and calculate the first review value of any review data. The calculation formula is:

[0096]

[0097] Among them, S1 is the first review value of any review data, W k is the weight of the kth second-category keyword, N6 is the number of sentences in any review data, and N k is the ranking of the sentence where the kth second-category keyword is located, K1 is the total number of second-category keywords in any review data, and the second-category keywords are positive review words.

[0098] Step 24, traverse all the review data in the first historical review data set. When the same first-category keyword exists, take the average value of the first review value corresponding to the same first-category keyword as the first final review value of the same first-category keyword, and then store the first user identifier, any first-category keyword and the first final review value corresponding to any first-category keyword in the second storage module accordingly.

[0099] Step 25: extract the second user identifier of any second user in the second historical review data set, obtain browsing data corresponding to any review data of any second user in the second historical review data set, and calculate the second final review value corresponding to any first-category keyword by any second user according to the methods of steps 22 to 24.

[0100] Step 26: After traversing all the second users in the second historical review data set, any second user identifier, any first category keyword and the second final review value corresponding to any first category keyword are correspondingly stored in the second storage module.

[0101] Step 27: define the first category of keywords corresponding to the first final review value greater than the eighth preset value as keywords to be recommended, and calculate the similarity of the reviews of the first user and any second user on any keyword to be recommended based on the first final review value of any keyword to be recommended by the first user and the second final review value of any keyword to be recommended by any second user; when the review similarity is greater than the ninth preset value, classify the first user and any second user into the same user set.

[0102] Preferably, a technology service related vocabulary set is set in advance, keywords of specific topics / review data are extracted, and it is determined whether the above keywords belong to the technology service related vocabulary set. If so, the keyword is a vocabulary related to technology services. Exemplarily, the technology service related vocabulary set includes technology service fields and / or technology service providers.

[0103] Preferably, a weight is set in advance for each second category keyword. For example, the weight of "like" is 7, and the weight of "very like" is 10.

[0104] For example, the review sentence is “..., .. like it very much, ....", the number of sentences in the review data is 3, and the rank of the sentence where “like it very much” is 2.

[0105] Preferably, the formula for calculating the review similarity is:

[0106]

[0107] Among them, SCMpq is the similarity of the comments between the first user and the qth second user on the pth keyword to be recommended, SF1p is the first final comment value of the first user on the pth keyword to be recommended, and SF2p is the second final comment value of the qth second user on the pth keyword to be recommended.

[0108] Exemplarily, the first final comment value of the first user for any keyword to be recommended is 8, the second final comment value of the second user for any keyword to be recommended is 9, and the similarity of the comments of the first user and the second user for any keyword to be recommended is 1-|8-9|÷8=0.875.

[0109] Users in the same user set have the same or similar preferences. Recommending scientific and technological service information that other users in the same user set are interested in to the first user can help the first user broaden his or her thinking and keep abreast of new trends in scientific and technological services.

[0110] Specifically, before acquiring the first historical review data set and the second historical review data set, a unique authentication identifier is allocated to any user according to the real-name authentication information of any user, and the historical review data of any user is collected based on the authentication identifier.

[0111] On different social platforms, users have different accounts and different user names. Therefore, before obtaining the user's historical review data, a unique authentication identifier is assigned to each user based on the user's real-name authentication information. In the historical review data set, the unique authentication identifier of any user is stored corresponding to the historical review data of any user.

[0112] Preferably, the above-mentioned social platform is a platform connected to the data recommendation terminal, and the historical review data obtained is more professional and has a strong relevance to the target data to be searched.

[0113] The push data is screened and sorted according to preset rules and sent to the user terminal, wherein the push data includes first push data and second push data.

[0114] Specifically, filtering and sorting the push data according to the preset rules and sending the push data to the user terminal includes:

[0115] Step 31, determine whether the first category of keywords in any push data title exists in the first storage module, if so, delete any push data, and after traversing all push data, define the remaining push data as data to be pushed.

[0116] Step 32: define the first category of keywords in any title of data to be pushed as keywords to be sorted, and determine whether the gth keyword to be sorted exists in the second storage module. If so, proceed to step 33; if not, proceed to step 34, where g is a positive integer from 1 to G, and G is the total number of keywords to be sorted.

[0117] Step 33: Based on the second final comment value of any second user on the g-th keyword to be sorted, and the similarity of comments between any second user and the first user on the g-th keyword to be sorted, the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended is calculated, and the calculation formula is:

[0118]

[0119] Among them, RL g is the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended, SC gq C2 is the similarity between the comments of the qth second user and the first user on the gth keyword to be recommended, gq is the second final comment value of the qth second user on the gth keyword to be recommended, and Q is the total number of second users who comment on the gth keyword to be recommended.

[0120] Step 34: Obtain the click rate of the data to be pushed corresponding to the g-th keyword to be sorted.

[0121] Step 35: After traversing all the keywords to be sorted, sort the data to be pushed based on the recommendation degree and / or click rate of the data to be pushed.

[0122] Based on the same user set, the second final comment value of each second user for the keyword to be sorted, and the similarity of the comments of each second user and the first user on the keyword to be recommended, the recommendation degree of the keyword to be recommended (that is, the recommendation degree of the data to be pushed corresponding to the keyword to be recommended) is calculated, and the data to be pushed is sorted and displayed based on the recommendation degree.

[0123] Exemplarily, the first push data is displayed first, the data to be pushed in the first push data is sorted based on the recommendation degree, the data to be pushed without recommendation degree is sorted according to the click rate, and then the data to be pushed in the second push data is sorted based on the recommendation degree.

[0124] Figure 2 FIG. 1 is a schematic diagram of a structure of an embodiment of a data push system provided by the present invention. Figure 2 As shown, the system includes: a data push terminal 100 and a user terminal 200 , and the data push terminal 100 includes a data receiving module 101 , a first push module 102 , a second push module 103 , and a data output module 104 .

[0125] The data receiving module 101 is used to obtain a first search content input by a first user.

[0126] The first push module 102 is used to traverse the first database according to the first search content, obtain a first question statement whose similarity with the first search content is greater than a first preset value, further obtain N1 first reply statements corresponding to the first question statement in the first database, cluster the N1 first reply statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to the user terminal 200 of the first user, receive the second search content input by the first user according to the prompt information, and then search according to the second search content to obtain the first push data.

[0127] The second push module 103 is used to obtain the first user identifier of the first user, determine whether the first user is in any user set according to the first user identifier, and if so, obtain the first search records of all users in any user set within the first preset time, and use the browsing data with a click rate higher than a second preset value in the first search record as the second push data; if not, obtain the first historical comment data set of the first user, calculate the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, and divide the first user into the corresponding user set based on the comment similarity, and then obtain the second search records of all users in the corresponding user set within the first preset time, and use the browsing data with a click rate higher than the second preset value in the second search record as the second push data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is the browsing data commented on by the first user, and the second user is a user other than the first user.

[0128] The data output module 104 is used to filter and sort the push data according to preset rules and send the push data to the user terminal 200, wherein the push data includes first push data and second push data.

[0129] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0130] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0131] The above-mentioned embodiments only express the preferred implementation mode of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A data push method, characterized in that: The steps include: Obtain a first search content input by a first user, traverse a first database based on the first search content, obtain a first question statement whose similarity to the first search content is greater than a first preset value, further obtain N1 first answer statements corresponding to the first question statement in the first database, cluster the N1 first answer statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to a user terminal of the first user, receive a second search content input by the first user according to the prompt information, and then search according to the second search content to obtain first push data; Obtaining a first user identifier of the first user, determining whether the first user is in any user set according to the first user identifier, and if so, obtaining first search records of all users in any user set within a first preset time, and using browsing data in the first search records with a click rate higher than a second preset value as second pushed data; If not, then obtain a first historical comment data set of the first user, calculate the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, divide the first user into a corresponding user set based on the comment similarity, and then obtain second search records of all users in the corresponding user set within the first preset time, and use browsing data with a click rate higher than the second preset value in the second search record as the second pushed data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical comment data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is browsing data commented on by the first user, and the second user is a user other than the first user; Filtering and sorting the push data according to preset rules and sending the push data to the user terminal, wherein the push data includes the first push data and the second push data; The obtaining of a first historical review data set of the first user, calculating the review similarity between the first user and the second user based on the first historical review data set and the second historical review data set, and dividing the first user into a corresponding user set based on the review similarity comprises: Step 21: Obtain browsing data corresponding to any review data in the first historical review data set; Step 22: determine whether the browsing data corresponding to any of the review data is a discussion on a specific topic or a free discussion. If it is a discussion on a specific topic, extract keywords of the specific topic as first-category keywords of any of the review data. If it is a free discussion, extract first-category keywords corresponding to any of the review data from any of the review data, wherein the first-category keywords are words related to technology services. Step 23, perform semantic analysis on any of the review data to determine whether any of the review data is a positive review or a negative review. If it is a negative review, store the first category keywords corresponding to any of the review data in the first storage module. If it is a positive review, extract all the second category keywords in any of the review data and calculate the first review value of any of the review data. The calculation formula is: , Wherein, S1 is the first review value of any review data, W k is the weight of the kth second-category keyword, N6 is the number of sentences in any of the review data, N k is the ranking of the sentences where the kth second-category keyword is located, K1 is the total number of the second-category keywords in any of the review data, and the second-category keywords are positive review words; Step 24: traverse all review data in the first historical review data set, and when there are identical first-category keywords, use the average of the first review values ​​corresponding to the identical first-category keywords as the first final review value of the identical first-category keywords, and then store the first user identifier, any first-category keyword, and the first final review value corresponding to any first-category keyword in the second storage module in a corresponding manner; Step 25: extract the second user identifier of any second user in the second historical review data set, obtain browsing data corresponding to any review data of any second user in the second historical review data set, and calculate the second final review value corresponding to any first-category keyword by any second user according to the method of steps 22 to 24; Step 26: After traversing all second users in the second historical review data set, any second user identifier, any first-category keyword, and any second final review value corresponding to the first-category keyword are stored in the second storage module in a corresponding manner; Step 27: Define the first category of keywords corresponding to the first final review value greater than the eighth preset value as keywords to be recommended; calculate the review similarity of the first user and any of the second users on any of the keywords to be recommended based on the first final review value of the first user and the second final review value of any of the keywords to be recommended by the second user; and when the review similarity is greater than a ninth preset value, classify the first user and any of the second users into the same user set.

2. A data push method according to claim 1, characterized in that: The step of establishing the first database comprises: Step 111: traverse the social platforms connected to the data recommendation terminal to obtain search information input by any user, and if the search information meets the first preset condition, define the search information as a question statement to be selected; Step 112: Obtain all answer information corresponding to the question statement to be selected, and if any answer information meets the second preset condition, define any answer information as the first answer statement to be selected; Step 113: Calculate the answer similarity between any two first answer statements to be selected. If the answer similarity is greater than a fourth preset value, delete any one of the two first answer statements to be selected. After traversing all first answer statements to be selected, define the remaining first answer statements to be selected as answer statements to be selected. Step 114: after traversing all social platforms, the question similarity between any two question statements to be selected is calculated. If the question similarity is greater than a fifth preset value, any one of the two question statements to be selected is deleted, and the answer statements to be selected corresponding to the two question statements to be selected are merged. After the merging is completed, the duplicate answer statements to be selected are deleted. Step 115: After traversing all the question statements to be selected, define the remaining answer statements to be selected as question statements, and define the answer statements to be selected corresponding to the question statements as answer statements, and then store the question statements and the answer statements corresponding to the question statements in correspondence to generate the first database.

3. A data push method according to claim 2, characterized in that: The first preset condition is that the search information contains a question word; The second preset condition is that the similarity between any of the answer information and the question statement to be selected is less than a sixth preset value and any of the answer information is a supplement to the question statement to be selected.

4. A data push method according to claim 1, characterized in that: The clustering of the N1 first reply statements and the semantic analysis of the clustering results to generate prompt information comprises the following steps: Step 121: Calculate the first similarity between the first search content and any first reply statement. If the first similarity is greater than a seventh preset value, delete any first reply statement. After traversing all first reply statements, define the remaining first reply statements as second reply statements. Step 122: cluster the second reply statements to generate N2 reply clusters; Step 123: Use N3 decision trees to score any second reply statement, and use the average of the score values ​​of all decision trees as the statement score of any second reply statement; Step 124: Calculate the cluster score of any of the reply clusters based on the sentence scores of all second reply sentences in any of the reply clusters. The calculation formula is: , Among them, CS i is the clustering score of the i-th answer cluster, N4 is the total number of the second answer statements, M i is the number of second reply statements in the i-th reply cluster, R j is the sentence score of the j-th second reply sentence in the i-th reply cluster; Step 125: sort all reply clusters based on the cluster scores, perform semantic analysis on the top N5 reply clusters, generate N5 prompt messages and send them to the user terminal.

5. A data push method according to claim 1, characterized in that: Before acquiring the first historical review data set and the second historical review data set, a unique authentication identifier is allocated to any user according to the real-name authentication information of any user, and the historical review data of any user is collected based on the authentication identifier.

6. A data push method according to claim 1, characterized in that: The filtering and sorting of the push data according to the preset rules and sending the push data to the user terminal comprises: Step 31, determining whether the first category of keywords in any push data title exists in the first storage module, if so, deleting any of the push data, and after traversing all the push data, defining the remaining push data as data to be pushed; Step 32: define the first category of keywords in any title of the data to be pushed as keywords to be sorted, and determine whether the gth keyword to be sorted exists in the second storage module. If so, proceed to step 33; if not, proceed to step 34, wherein g is a positive integer from 1 to G, and G is the total number of keywords to be sorted; Step 33: Based on the second final comment value of any second user on the g-th keyword to be sorted, and the similarity of comments between any second user and the first user on the g-th keyword to be sorted, the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended is calculated, and the calculation formula is: , Among them, RL g is the recommendation degree of the data to be pushed corresponding to the g-th keyword to be recommended, SC gq C2 is the similarity between the comments of the qth second user and the first user on the gth keyword to be recommended, gq is the second final comment value of the qth second user on the gth keyword to be recommended, and Q is the total number of second users who comment on the gth keyword to be recommended; Step 34: Obtain the click rate of the data to be pushed corresponding to the g-th keyword to be sorted; Step 35: After traversing all the keywords to be sorted, sort the data to be pushed based on the recommendation degree and / or click rate of the data to be pushed.

7. A data push system, used to implement the data push method according to any one of claims 1 to 6, characterized in that: include: A data push terminal and a user terminal, wherein the data push terminal includes a data receiving module, a first push module, a second push module, and a data output module; The data receiving module is used to obtain a first search content input by a first user; The first push module is used to traverse the first database according to the first search content, obtain a first question statement whose similarity with the first search content is greater than a first preset value, further obtain N1 first answer statements corresponding to the first question statement in the first database, cluster the N1 first answer statements, perform semantic analysis on the clustering results to generate prompt information, and send the prompt information to the user terminal of the first user, receive the second search content input by the first user according to the prompt information, and then search according to the second search content to obtain the first push data; The second push module is used to obtain a first user identifier of the first user, determine whether the first user is in any user set according to the first user identifier, and if so, obtain first search records of all users in any user set within a first preset time, and use browsing data in the first search record with a click rate higher than a second preset value as second push data; If not, then obtain a first historical comment data set of the first user, calculate the comment similarity between the first user and the second user based on the first historical comment data set and the second historical comment data set, divide the first user into a corresponding user set based on the comment similarity, and then obtain second search records of all users in the corresponding user set within the first preset time, and use browsing data with a click rate higher than the second preset value in the second search record as the second pushed data, wherein the first historical comment data set is a set of comment data published by the first user within the second preset time, the second historical comment data set is a set of comment data published by the second user for each first browsing data within the second preset time, the first browsing data is browsing data commented on by the first user, and the second user is a user other than the first user; The data output module is used to filter and sort the push data according to a preset rule and send the push data to the user terminal, wherein the push data includes the first push data and the second push data; In the second push module, the obtaining of a first historical review data set of the first user, calculating the review similarity between the first user and the second user based on the first historical review data set and the second historical review data set, and dividing the first user into a corresponding user set based on the review similarity includes: Step 21: Obtain browsing data corresponding to any review data in the first historical review data set; Step 22: determine whether the browsing data corresponding to any of the review data is a discussion on a specific topic or a free discussion. If it is a discussion on a specific topic, extract keywords of the specific topic as first-category keywords of any of the review data. If it is a free discussion, extract first-category keywords corresponding to any of the review data from any of the review data, wherein the first-category keywords are words related to technology services. Step 23, perform semantic analysis on any of the review data to determine whether any of the review data is a positive review or a negative review. If it is a negative review, store the first category keywords corresponding to any of the review data in the first storage module. If it is a positive review, extract all the second category keywords in any of the review data and calculate the first review value of any of the review data. The calculation formula is: , Wherein, S1 is the first review value of any review data, W k is the weight of the kth second-category keyword, N6 is the number of sentences in any of the review data, N k is the ranking of the sentences where the kth second-category keyword is located, K1 is the total number of the second-category keywords in any of the review data, and the second-category keywords are positive review words; Step 24: traverse all review data in the first historical review data set, and when there are identical first-category keywords, use the average of the first review values ​​corresponding to the identical first-category keywords as the first final review value of the identical first-category keywords, and then store the first user identifier, any first-category keyword, and the first final review value corresponding to any first-category keyword in the second storage module in a corresponding manner; Step 25: extract the second user identifier of any second user in the second historical review data set, obtain browsing data corresponding to any review data of any second user in the second historical review data set, and calculate the second final review value corresponding to any first-category keyword by any second user according to the method of steps 22 to 24; Step 26: After traversing all second users in the second historical review data set, any second user identifier, any first-category keyword, and any second final review value corresponding to the first-category keyword are stored in the second storage module in a corresponding manner; Step 27: Define the first category of keywords corresponding to the first final review value greater than the eighth preset value as keywords to be recommended; calculate the review similarity of the first user and any of the second users on any of the keywords to be recommended based on the first final review value of the first user and the second final review value of any of the keywords to be recommended by the second user; and when the review similarity is greater than a ninth preset value, classify the first user and any of the second users into the same user set.