Expert recommendation method based on multi-source information fusion technology

Abstract

The invention discloses an expert recommendation method based on a multi-source information fusion technology, and the method comprises the steps: crawling technical expert science and technology papers, invention patents, fund project information and Web page information to construct a knowledge base, and constructing a keyword dictionary keyworks according to a keyword field of the knowledge base; extracting knowledge base author fields to carry out word frequency co-occurrence analysis, and constructing an expert cooperation relationship subnet; extracting a Web page expert research direction and personal information by using a regular expression and a named entity recognition algorithm respectively to construct a Web subnet; extracting document topics and topic keywords from abstract fields of the knowledge base through an lda algorithm, and extracting five words with the maximum abstract field weight through a TFIDF algorithm to jointly construct a topic subnet; and constructing and calculating expert centrality values in the expert information network by taking an expert name mechanism as a constraint condition and combining the three sub-networks, sorting the expert centrality values, and recommending the top five experts as recommendation results.

Description

technical field [0001] The invention belongs to the field of multi-source information fusion and expert recommendation, in particular to an expert recommendation method based on multi-source information fusion technology. Background technique [0002] Traditional technical expert recommendation algorithms usually use a single data source for recommendation. The single data source recommendation is easily constrained by the data source, and it is easy to cause the lack of expert information, so that the expert information cannot be fully displayed and the expert recommendation is relatively isolated. If the relationship with the working institution cannot be effectively expanded, researchers can only recommend expert information with one attribute, while the multi-source information fusion method can integrate the multi-dimensional attributes of experts according to the constraints, so as to fully display the technical expert information, and The expert cooperation relationsh...

AI Technical Summary

Problems solved by technology

[0007] Existing expert recommendation patent applications include: Su Yurong, Li Zhenhua, an expert recommendation method and system: CN111160699 A, 2020.05.15, although it can be implemented on the basis of multiple recommendation systems to give users more standard recommendation results, there may be The problem of information redundancy, and the problem of missing information cannot be avoided; Zhang Yongfeng, Tan Xinyu, Li Zhenhua, an expert recommendation method and system based on multiple data sources: CN 111008330 A, 2020.04.14, by adding the score field to the expert and according to the score field The corresponding value sorts the experts and generates recommendation results. Although the patent involves multi-source data, it only scores the expert fields and fails to mine the hidden expert relationship between multiple data; Wang Jian, Sun Jiqing, Lin Hongfei, a A community Q&A expert recommendation method based on a recurrent neural network: CN 108021616 A, 2018.05.11. This recommendation method can effectively represent sentence grammar and semantic information, reducing manual intervention on recommendation results, but it is easily affected by the original corpus and thus Recommended results have an impact

[0008] Although the above patents have effectively improved the recommendation results, they do not involve the recommendation of expert relationships and regional relationships. They only recommend from the experts themselves, and cannot comprehensively consider factors such as regional relationships and cooperative relationships, resulting in invalid recommendation results and cannot be applied. to actual recommendation

Images