A Paper Recommendation Method Based on Distributed Expression of Citation Relationship

Abstract

The invention discloses a rational distributed expression reference based thesis recommendation method. Context of theses in weighting reference network is expressed by the aid of distributed vectors, similarity among the theses is calculated by the aid of the vectors, and the purpose of thesis recommendation is achieved; previous reference based thesis recommendation methods are confined to using of contact ratio of a thesis collection citing and cited by the theses, and the similarity among the theses with the contact ratio being zero cannot be calculated. Through the weighting reference network among the theses, 'indirect reference' information among the theses are fully utilized, distributed vectors used for expressing locations of the theses in the reference network are acquired with a matrix splitting method, and inner product is taken as the similarity expression among the theses; the distributed expression can be taken as a datum feature for an academic recommendation system, problems existing in an existing model can be well solved, and correctness in similarity measurement can be further improved.

Description

technical field [0001] The invention relates to the technical field of recommendation, in particular to a method for recommending papers based on the distributed expression of citation relationships. Background technique [0002] As an important part of academic services, the academic recommendation system has been introduced into many academic search engine systems, such as CiteSeerX, Google Scholar, PubMed and so on. Through the academic recommendation system, researchers will be recommended other papers related to their existing research papers. Researchers can check the integrity of their reference materials, save manpower in searching for related jobs, and quickly access new papers in their field. [0003] The essence of the recommendation system is to guess and push specific items of interest for specific users. According to the information-based recommendation, it is mainly divided into content-based recommendation, collaborative filtering-based recommendation, socia...

AI Technical Summary

Problems solved by technology

[0004] The existing work of providing recommendation information based on paper citation information, the representative models include the traditional reference prediction model and collaborative filtering model, which are limited to the use of the coincidence degree of the paper citation and the cited paper collection, for the coincidence degree is 0 The similarity cannot be calculated between the papers of

while ignoring available information on "indirect citations" between papers

Images