Recommender resources based on acquiring user's requirement and exploring user's preference with Word2Vec model in web service

摘要

Traditional recommender algorithms mainly use structured data (resource tag, user feature etc.) to depict the user preference and ignore the semantic relations of resources. In this paper, we present a new idea for acquiring user's requirement and exploring user's preference with Word2Vec model (RP-Word2Vec) to find the interested and personal resource in the web service. We use Word2Vec model to measure the sentiment among keywords and acquire user's requirement as accurately as possible; and we treat resources as the input of Word2Vec model based on history behaviours and adopt a semantic similarity measuring process to recommend interested and personal resource for the user. The experiments results that the presented RP-Word2Vec supports more effective.