TWI845796B - Knowledge graph association search method and system - Google Patents

Abstract

A knowledge graph association search method has steps of: receiving a key word input by a user via a user terminal device; perform data pre-processing on the key word; acquiring a key word model; based upon a keyword model, obtaining similar key words being similar to the pre-processed key word; searching an article/knowledge point word/score database according to the pre-processed key word and the similar key words, so as to obtain a searching result, wherein the searching result comprises associated words, associated articles and associated knowledge of the pre-processed key word and the similar key words, and the associated words are classified by knowledge point words; and, according to the searching result, presenting a page of the knowledge point graph, the associated articles and the associated knowledge to the user.

Description

Knowledge graph associative search method and system thereof

The present invention relates to a search method and system that can generate search results based on keywords, and in particular to a knowledge graph associative search method and system that presents the search results in the form of knowledge point graphs, related knowledge and related articles.

A knowledge point graph is a data structure represented by a diagram, which has a root node and multiple nodes and connecting lines under the root node, wherein the root node and multiple nodes are connected by connecting lines, and the words in the connected nodes have an associative relationship. The topology of the knowledge point graph can be a tree, which classifies the associative words of multiple nodes under the root node according to the knowledge point words. Through the presentation of the knowledge point graph, users can understand the associative relationship between the knowledge point words, the words of the root node, and the associative words.

Existing search systems can search for relevant web pages, documents or multimedia files based on keywords or pictures entered by users. However, for users who want to learn knowledge, the search results found through existing search systems may not effectively help users learn, and for users without relevant background knowledge, it is not easy to achieve the purpose of effective and rapid learning because they do not know other associated words corresponding to keywords, or even do not know how to classify the associated words of keywords. For example, for users who have never invested in stocks, if they simply search for "stocks" through the search system, it is not easy for users to effectively and quickly learn stock investment knowledge because they do not know other associated words of "stocks" (for example, words such as listing, over-the-counter, outstanding stocks and dividends).

The present invention provides a knowledge graph association search method and system that presents search results in the form of knowledge point graph, related knowledge and related articles. The knowledge graph association search method and system can present a page of search results of the knowledge point graph, related knowledge and related articles with the keyword as the root node to the user according to the keyword (i.e., the word of the root node) input by the user. Through the knowledge point map, users can know the words associated with the keywords they input based on the knowledge point words, and read the related knowledge and related articles of the associated words to quickly and effectively learn the knowledge corresponding to the keywords. The knowledge point words are preset by the search service provider, and can be, for example, "business thinking", "industry thinking" and "investment thinking".

Based on one of the objectives to be achieved and the technical problems to be solved by the present invention, an embodiment of the present invention provides a knowledge graph associative search method, which is executed on a knowledge graph associative search server, and includes: receiving a keyword word input by a user through a user device; performing data pre-processing on the keyword word; obtaining a keyword word model; obtaining a plurality of similar keyword words similar to the pre-processed keyword word based on the keyword word model; and obtaining a plurality of similar keyword words according to the pre-processed keyword word model. A keyword and the multiple similar keyword search articles/knowledge point words/score database are used to obtain search results, wherein the search results include multiple associated words, multiple related articles and related knowledge related to the keyword and the multiple similar keyword, and the associated words are classified according to the multiple knowledge point words; and based on the search results, a knowledge point map, the multiple related articles and the related knowledge pages are presented to the user.

According to the aforementioned technical features, the plurality of knowledge point words are respectively "business thinking", "industry thinking" and "investment thinking".

According to the aforementioned technical features, the article/knowledge point word/score database records the article identification, knowledge point word, item name and item score corresponding to the article in the article database.

According to the aforementioned technical features, the article database includes an internal article database and an external article database.

According to the aforementioned technical features, a search is performed in the article/knowledge point word/score database to obtain a plurality of knowledge point words and item names identified by articles corresponding to the keyword words, a plurality of knowledge point words and item names identified by articles corresponding to the keyword words are merged, and the knowledge point graph is generated based on the merged plurality of knowledge point words and the plurality of item names, wherein the merged plurality of item names serve as the plurality of association words of the knowledge point graph.

According to the aforementioned technical features, the root node of the knowledge point graph is used to present the keyword, and the multiple root nodes under the root node presenting the multiple associated words can be selected by the user to present the multiple related articles and the related knowledge associated with the associated words and the keyword.

Based on one of the objectives to be achieved and the technical problems to be solved by the present invention, an embodiment of the present invention provides a knowledge graph associative search system, which includes: a knowledge graph associative search server; and a user device of at least one user, wherein the knowledge graph associative search server is used to execute the above-mentioned knowledge graph associative search method.

In short, through the knowledge graph associative search method and system provided by the embodiment of the present invention, even if the user does not understand other associative words corresponding to the keyword, he can still learn the related associative words through the present invention, and learn the related knowledge through the article database provided by the knowledge graph associative search method and system. In this way, it is beneficial for the user to learn the related knowledge quickly and effectively.

In order to fully understand the purpose, features and effects of the present invention, the present invention is described in detail through the following specific embodiments and the accompanying drawings.

The embodiment of the present invention provides a knowledge graph associative search system in which search results are presented in the form of knowledge point graph, related articles and related knowledge. The knowledge graph associative search system includes at least one user device and a knowledge graph associative search server, wherein the user device and the knowledge graph associative search server are connected to each other in communication, the knowledge graph associative search server can be implemented by hardware circuits and software, or all by hardware circuits, the user device can be a smart phone, a laptop, a desktop computer, a tablet computer or a netbook, and the present invention is not limited thereto.

The user can send the input keyword to the knowledge graph association search server through the user device, so that the knowledge graph association search server draws a knowledge point graph with the keyword as the root node according to the keyword, and presents the knowledge point graph to the user as at least a part of the search results, wherein the association words of the multiple nodes under the root node of the knowledge point graph are classified by multiple knowledge point words, which are preset by the search service provider and can be, for example, "business thinking", "industry thinking" and "investment thinking". In addition, the associated knowledge and related articles of the association words are also presented to the user as part of the search results. Furthermore, in the embodiment of the present invention, the knowledge graph association search server searches the article/knowledge point word/score database to obtain multiple association words, related knowledge of the association words, and related articles of the association words, wherein the article/knowledge point word/score database is obtained by processing data from internal and external article databases.

Since there may be a lot of associative words associated with keywords, the associative words are limited to those classified by knowledge point words in the article/knowledge point word/score database, and the associated knowledge and related articles can be stored in the internal article database and the external article database, so that users can read the associated knowledge and related articles through the page presenting the search results, so as to achieve the purpose of learning knowledge quickly and effectively. The knowledge point words in the article/knowledge point word/score database are defined by the operator of the knowledge graph associative search system, which usually hopes to let users learn the knowledge associated with the keyword. For example, if the user is expected to learn relevant investment knowledge through the knowledge graph associative search system, the knowledge point words can be preset to the above-mentioned "business thinking", "industry thinking" and "investment thinking". For another example, if the user is expected to learn relevant semiconductor technology through the knowledge graph associative search system, the knowledge point words can be modified to "semiconductor materials", "semiconductor physics" and "semiconductor process". In short, the setting of knowledge point words is not a limitation of the present invention.

The association words are the item names of the knowledge point words in the article/knowledge point word/score database. The item names are obtained through article analysis, such as using the vocabulary or item name/knowledge point word database to map the non-relational article database (such as the Elastic Search (ES) article database). In addition, since the number of secondary association words corresponding to the knowledge point words (that is, the item names searched in the article/knowledge point word/score database) may also be quite large, the item scores in the article/knowledge point word/score database can be used as a criterion for whether to be presented in the knowledge point map. For example, if there are 15 association words corresponding to a certain knowledge point word corresponding to the input keyword in the article/knowledge point word/score database, the item scores of these 15 association words can be counted, and the 6 association words with higher item scores can be presented in the knowledge point map (but the present invention is not limited to presenting the 6 association words with higher item scores in the knowledge point map, and the number can be changed according to the actual situation). In addition, based on the above-mentioned knowledge graph association search system, the embodiment of the present invention provides a corresponding knowledge graph association search method.

After explaining the inventive concept of the knowledge graph associative search system and method of the embodiment of the present invention, the details of the knowledge graph associative search system and method of the embodiment of the present invention are further explained. First, please refer to Figure 1, which is a block diagram of the knowledge graph associative search system of the embodiment of the present invention. The knowledge graph associative search system 1 includes a knowledge graph associative search server 11 and at least one user device 12, wherein the knowledge graph associative search server 11 is communicatively connected to the user device 12, and can also be communicatively connected to an external server 13. The knowledge graph association search server 11 can be implemented by hardware circuits and software, or all by hardware circuits. The user device 12 can be a smart phone, a laptop, a desktop computer, a tablet computer or a netbook, and the present invention is not limited thereto.

The knowledge graph associative search server 11 receives keywords from the user device 12 and generates search results based on the keywords. The knowledge graph associative search server presents a knowledge point graph with the keywords as the root node based on the search results, and the multiple associative words under the root node of the knowledge point graph are classified according to the knowledge point words. For example, the multiple associative words under the root node are classified into "business thinking", "industry thinking" and "investment thinking", and the associative words of each category are connected to the root node through a connecting line.

Next, the implementation of the knowledge graph associative search server 11 is further described, but the present invention is not limited to the implementation of the knowledge graph associative search server 11 described below. Please refer to FIG. 2, which is a block diagram of the knowledge graph associative search server of an embodiment of the present invention. In FIG. 2, the knowledge graph associative search server 11 includes a processing unit 21, an input/output interface unit 22, a storage unit 23, and a memory 24, wherein the processing unit 21 is electrically connected to the input/output interface unit 22, the storage unit 23, and the memory 24.

In this embodiment, the processing unit 21 is used to execute the software program to perform the knowledge graph associative search. The input and output interface unit 22 includes a communication module to enable the knowledge graph associative search server 11 to communicate with the user device, thereby receiving the keywords of the user device and transmitting the search results to the user device. The storage unit 23 stores the aforementioned software program, and can also be planned with a specific database unit 231 to store various types of databases. The memory 24 serves as a data storage space when the processing unit 21 executes the software program. It should be noted that in other embodiments, the storage unit 23 may not have a specific part for storing various databases, or various databases may be stored outside the knowledge graph association search server 11. In short, the present invention is not limited by the storage method of various databases.

Next, please refer to Figure 3, which is a flow chart of the knowledge graph associative search method of an embodiment of the present invention. The knowledge graph associative search method can be executed by the above-mentioned knowledge graph associative search server 11, and each step is described as follows. First, in step S31, the keyword word input by the user through the user device is received. Then, in step S32, the keyword word is pre-processed, and the data pre-processing here includes correcting the spelling errors and semantic analysis of the keyword word to obtain the word vector of the keyword word. Then, in step S33, a keyword word model is obtained, wherein the keyword word model records the word vector of each word. Then, in step S34, similar keywords similar to the pre-processed keywords are obtained based on the keyword model, and by calculating the distance between the keyword vector and the word vector of each word in the keyword model, similar keywords similar to the pre-processed keywords can be found.

Next, in step S35, the article/knowledge point word/score database is searched based on the pre-processed keywords and similar keywords to obtain search results, wherein the search results include associated words, related articles and related knowledge related to the keywords and similar keywords, and the associated words are classified by knowledge point words. Further, the knowledge graph association search server searches the article/knowledge point word/score database to obtain multiple associated words, associated knowledge of the associated words and associated articles of the associated words. Then, in step S36, based on the search results, the knowledge point map, related articles and related knowledge pages are presented to the user, that is, the above pages are generated for the user device based on the search results, wherein the root node of the knowledge point map has keywords, and the associated words of the multiple nodes under the root node are classified by multiple knowledge point words.

Next, please refer to FIG. 4, which is a flow chart of obtaining an article/knowledge point word/score database in an embodiment of the present invention. FIG. 4 illustrates a method of obtaining an article/knowledge point word/score database in step S35 of FIG. 3, which is not intended to limit the present invention. In step S41, the knowledge graph association search server first obtains internal and external article databases (it may also obtain only the internal article database) and the item name/knowledge point word database. The internal article database records multiple articles after screening. These multiple articles are screened by the operators of the knowledge graph association search system manually or automatically by software programs. The purpose is to allow users to learn the knowledge that the operators want users to learn through these articles. The external article database records multiple unscreened articles to avoid analyzing too many articles. It can be designed to analyze only articles that are popular or have a large number of readers before a specific date. By the way, the multiple articles recorded in the internal and external article databases can include text, pictures, audio and video and/or keyword content, etc., that is, the article database is not limited to recording only text content, and the article itself can even be just pictures or audio and video. The item name/knowledge point word database records the relationship between multiple knowledge point words and item names preset by the operator. Part of the content of the item name/knowledge point word database can be shown in Table 1.

Table 1: Partial content of the project name/knowledge point word database Project Name Knowledge Points Chinese words English words Business Thinking Business Thinking Investment Thinking Investment Thinking Investment Thinking Industry Thinking Warren Buffett Warrant Buffet Punch Cards Moat ROE Free cash flow per share EPS Fast Food Business Opportunities Philip Fisher Philip Fisher Safety Margin Management Profitability Price to Book Ratio Net profit growth rate - McDonald's McDonalds Drive-thru Fast food reviews High-Performance Stocks Long-term investment - McDonald's

In step S42, the knowledge graph association search server performs article analysis on the articles in the internal and external article databases based on the item name/knowledge point word database. Then, in step S43, the knowledge graph association search server generates an article/knowledge point word/score database based on the analysis results. The article/knowledge point word/score database records the article identification, knowledge point word, item name and item score corresponding to the article in the article database, and preferably has a topic type, topic name and topic score, wherein part of the content of the article/knowledge point word/score database can be shown in Table 2.

Table 2: Partial content of the article/knowledge point word/score database Article Identification Subject Category Theme Name Topic score Knowledge Points Project Name Item score 30461 Wafer Foundry TSMC 13.25 Business Thinking Morris Chang 4.12 Industry Thinking Wafer Foundry 5.11 Investment Thinking Value Stocks 4.09 59018 Corporate Governance Value Stocks 9.29 Business Thinking Corporate Governance 6.24 Industry Thinking Wafer Foundry 3.59 Investment Thinking Three major legal persons 2.65

In addition, the method of conducting article analysis is roughly as follows. First, use the item name in the item name/knowledge point word database to search for articles in the article database, and determine the article's topic name and topic score based on the relevance between the article and the item name. Then, use each item name of the knowledge point word in the item name/knowledge point word database to search for articles in the article database, and determine the knowledge point word and the item name under it and its item score based on the relevance between the article and the item name of the knowledge point word. For example, a non-relational search (elastic search) scoring system can be used to calculate the frequency of each item name in each article and the total frequency of all articles to determine the item name of the knowledge point words of the article and its item score, but the present invention is not limited to this, and other search scoring systems can also be used to calculate and determine the item name of the knowledge point words of the article and its item score. In this way, each article has a corresponding topic name, topic score, knowledge point words, item name and item score, thereby forming an article/knowledge point word/score database as shown in Table 2. For example, for the article with article identification "30461", the item name "wafer foundry" appears most frequently (wherein the aforementioned subject score is related to this term frequency), therefore, the subject name of the article is determined to be "wafer foundry", and the term frequency of the item name "Zhang Zhongmou" in the knowledge point word "business thinking" exceeds a specific value (wherein the aforementioned item score is related to this term frequency), therefore, the item name under the knowledge point word "business thinking" of the article with article identification "30461" includes "Zhang Zhongmou".

Incidentally, although the above embodiment uses the item name/knowledge point word database to perform article analysis on the articles in the article database, the present invention is not limited thereto. In other embodiments, a word database may be used to perform article analysis on the articles in the article database. Furthermore, the above article analysis method is not intended to limit the present invention, and other methods that can perform article analysis on articles to determine the item names and knowledge point words of the articles can also be used in the present invention.

Next, please refer to FIG. 5, which is a flow chart of generating a knowledge point graph according to a keyword in an embodiment of the present invention. In step S51, the knowledge graph association search server searches the article database according to the keyword to obtain the article identification of the previous few articles (for example, 100 articles, but the present invention is not limited to this). In step S52, the knowledge graph association search server searches the article/knowledge point word/score database to obtain the item name of the knowledge point word corresponding to the article identification, for example, obtaining 100 knowledge point words and their item names of the article identification. Next, in step S53, the knowledge graph associative search server merges the knowledge point words and their item names of all articles obtained in step S52. Afterwards, in step S54, the knowledge graph associative search server draws the knowledge point graph of the corresponding keyword words according to the merged knowledge point words and their item names. In addition, in step S52, the item score corresponding to the item name identified in the article can be obtained as a criterion for whether the item name is merged. For example, an item name with a too low item score is discarded and will not appear in the drawn knowledge point graph. The remaining project names are presented as associated words in the drawn knowledge point map.

For example, if the article searched based on the keyword has the knowledge point word "industry thinking" and its item name "wafer foundry", the knowledge point word "business thinking" and its item name "Zhang Zhongmou", and the knowledge point word "investment thinking" and its item name "capital stocks", and another article searched based on the keyword has the knowledge point word "investment thinking" and its item name "three major laws "People" and the knowledge point word "industry thinking" and its item name "packaging", after merging, the item names of the keyword knowledge point word "industry thinking" include "wafer foundry" and "packaging", the item names of the keyword knowledge point word "investment thinking" include "capital stocks" and "three major legal persons", and the item name of the keyword knowledge point word "business thinking" includes "Zhang Zhongmou". Then, the knowledge point map drawn according to the above merging results includes the above knowledge point words and the item names under them.

Next, please refer to FIG. 6, which is a schematic diagram of the search results of the knowledge point graph generated based on the keyword in the embodiment of the present invention. In FIG. 6, the keyword input is "TSMC". After searching the article/knowledge point word/score database, the knowledge graph Lenovo search server found that several articles related to "TSMC" have high item scores for the item names "Mr. Chang", "Corporate Governance" and "Capital Expenditure" under the knowledge point word "Business Thinking", and several articles related to "TSMC" have high item scores for the knowledge point word "Industry". The item names "semiconductor", "packaging" and "wafer foundry" under "investment thinking" have higher item scores, and the item names "supply chain", "three major legal persons", "Taiwan 50", "stock" and "capital stocks" under the knowledge point words "investment thinking" in several articles related to "TSMC" have higher item scores, so after merging, the knowledge point map 61 drawn is as shown in Figure 6. In the page of Figure 6, the user selected the associated word "capital stocks", so the page of Figure 6 presents the related knowledge 62 of "TSMC" and "capital stocks", and the page of Figure 6 presents multiple related articles 63 of "TSMC" and "capital stocks".

In summary, the knowledge graph association search system and method of the embodiment of the present invention searches in the article/knowledge point word/score database according to the keyword to obtain the corresponding article identification, and obtains the item name of the knowledge point word of the article of the corresponding article identification as the association word, and merges the obtained association words of each knowledge point word, and draws a knowledge point graph with the keyword as the root node according to the merged result to present it to the user. Even if the user does not understand other association words corresponding to the keyword, he can learn the related association words through the present invention, and learn the related knowledge through the related articles and related knowledge provided by the knowledge graph association search system. This helps users learn relevant knowledge quickly and effectively.

The present invention has been disclosed in the above with preferred embodiments, but those skilled in the art should understand that the above embodiments are only used to describe the present invention and should not be interpreted as limiting the scope of the present invention. It should be noted that all changes and substitutions equivalent to the above embodiments should be included in the scope of the present invention. Therefore, the protection scope of the present invention shall be based on the scope defined by the application patent.

1: Knowledge graph associative search system 11: Knowledge graph associative search server 12: User device 13: External server 2: Knowledge graph associative search server 21: Processing unit 22: Input/output interface unit 23: Storage unit 231: Database unit 24: Memory 61: Knowledge point graph 62: Related knowledge 63: Related articles S31~S33, S41~S43, S51~S54: Steps

FIG1 is a block diagram of a knowledge graph associative search system according to an embodiment of the present invention.

FIG. 2 is a block diagram of a knowledge graph association search server according to an embodiment of the present invention.

FIG3 is a flow chart of the knowledge graph associative search method according to an embodiment of the present invention.

FIG. 4 is a flow chart of obtaining an article/knowledge point word/score database according to an embodiment of the present invention.

FIG5 is a flow chart of generating a knowledge point map based on keywords according to an embodiment of the present invention.

FIG6 is a schematic diagram of a knowledge point map, related articles, and search results of related knowledge generated based on keywords according to an embodiment of the present invention.

S31~S36: Steps

Claims (8)

A knowledge graph associative search method is executed on a knowledge graph associative search server, comprising: receiving a keyword word input by a user through a user device; performing data pre-processing on the keyword word; wherein the data pre-processing includes performing spelling correction and semantic analysis on the keyword word to obtain a word vector of the keyword word; obtaining a keyword word model; wherein the keyword word model records the word vector of each word; based on the keyword word The model obtains multiple similar keywords that are similar to the keyword after premenstrual processing; wherein the step of obtaining multiple similar keywords includes calculating the distance between the word vector of the keyword and the word vector of each word in the keyword model, and finding similar keywords that are similar to the keyword after premenstrual processing; searching the article/knowledge point word/score database according to the keyword after premenstrual processing and the multiple similar keywords to obtain a search result The method comprises the steps of: generating a search result, wherein the search result comprises a plurality of associated words associated with the keyword and the plurality of similar keywords, a plurality of associated articles associated with the plurality of associated words, and associated knowledge associated with the plurality of associated words; wherein the associated words are classified according to the plurality of knowledge point words, and the knowledge point words are predetermined; and presenting a page of a knowledge point map to the user according to the search result; receiving a request for the plurality of associated words; The selection of one or more of the associative words causes the page to simultaneously present the multiple related articles and the related knowledge pages associated with the selected associative words to the user, wherein the root node of the knowledge point graph has the keyword, and the multiple associative words of the multiple nodes under the root node are classified according to the multiple knowledge point words, and each category of the multiple associative words is connected to the root node through a connecting line. As in the knowledge graph associative search method of claim 1, the multiple knowledge point words are respectively "business thinking", "industry thinking" and "investment thinking". As in the knowledge graph associative search method of claim 1, the article/knowledge point word/score database records the article identification, knowledge point word, item name and item score of the article corresponding to the article database, and the article/knowledge point word/score database is obtained by a method comprising the following steps: obtaining an article database and an item name/knowledge point word database, wherein the item name/knowledge point word database records the relationship between a plurality of predetermined knowledge point words and item names; and performing article analysis on the articles in the article database based on the item name/knowledge point word database, and generating the article/knowledge point word/score database according to the analysis result. As in the knowledge graph associative search method of claim 1, the article database includes an internal article database and an external article database. As in claim 1, the knowledge graph associative search method is performed in the article/knowledge point word/score database to obtain multiple article-identified knowledge point words and item names corresponding to the keyword, multiple article-identified knowledge point words and item names corresponding to the keyword are merged, and the knowledge point graph is generated based on the merged multiple knowledge point words and the multiple item names, wherein the merged multiple item names are used as the multiple associative words of the knowledge point graph. A knowledge graph associative search system, comprising: a knowledge graph associative search server; and at least one user device; wherein the knowledge graph associative search server is used to perform the following steps: receiving a keyword word input by the user through the user device; performing data pre-processing on the keyword word; wherein the data pre-processing includes performing spelling correction and semantic analysis on the keyword word to obtain a word vector of the keyword word; obtaining a keyword word model; wherein the keyword word model records each The method comprises the steps of: obtaining a word vector of a word; obtaining a plurality of similar keywords similar to the keyword after premenstrual processing based on the keyword word model; wherein the step of obtaining a plurality of similar keywords comprises calculating the distance between the word vector of the keyword word and the word vector of each word in the keyword word model, and finding similar keywords similar to the keyword after premenstrual processing; searching an article/knowledge point word/score database according to the keyword after premenstrual processing and the plurality of similar keywords to obtain a search result, wherein the search The search results include multiple associated words related to the keyword and the multiple similar keyword, multiple related articles related to the multiple associated words, and related knowledge related to the multiple associated words; wherein the associated words are classified according to the multiple knowledge point words, and the knowledge point words are predetermined; and based on the search results, presenting a knowledge point map; receiving a selection of one or more of the multiple associated words, so that the page simultaneously presents the selected associated words. The user is provided with the pages of the multiple related articles and the related knowledge, wherein the root node of the knowledge point graph has the keyword, and the multiple associated words of the multiple nodes under the root node are classified by the multiple knowledge point words; The article/knowledge point word/score database records the article identification, knowledge point words, item name and item score corresponding to the article in the article database; wherein the multiple knowledge point words are respectively "business thinking", "industry thinking" and "investment thinking". As in claim 6, the knowledge graph associative search system, wherein the article/knowledge point word/score database is obtained by a method comprising the following steps: obtaining an article database and an item name/knowledge point word database, wherein the item name/knowledge point word database records the relationship between a plurality of predetermined knowledge point words and item names; and performing article analysis on the articles in the article database based on the item name/knowledge point word database, and generating the article/knowledge point word/score database based on the analysis results. As in claim 6, the knowledge graph associative search system is searched in the article/knowledge point word/score database to obtain multiple article-identified knowledge point words and item names corresponding to the keyword, multiple article-identified knowledge point words and item names corresponding to the keyword are merged, and the knowledge point graph is generated based on the merged multiple knowledge point words and the multiple item names, wherein the merged multiple item names are used as the multiple associative words of the knowledge point graph.