Open AccessA Text Mining Method to Trace Influential Product Features Based on PageRank
Author(s) -
Liu Qin,
Zhuchao Yu,
Shucheng He
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1288/1/012078
Subject(s) - computer science , pagerank , trace (psycholinguistics) , social media , similarity (geometry) , product (mathematics) , information retrieval , entropy (arrow of time) , feature (linguistics) , data science , data mining , world wide web , artificial intelligence , mathematics , linguistics , philosophy , physics , geometry , quantum mechanics , image (mathematics)
As a technology of Web2.0, social media has become a popular tool for both enterprises and customers to understand other voices. Therefore, an important project for enterprises to have an outstanding feedback is to use social media to promote and spread products correctly, so it’s necessary to mine some product features which customers care mostly. In this paper, we crawled online data from website DouBan to trace those features through their posted contents, such as descriptions, reviews, comments, etc. Firstly, we computed received attention of these contents by self-information algorithm and measured similarity by cross-entropy algorithm. Secondly, using the attention value as node weight and similarity as edge weight, we computed influence of product features in PageRank network by the proposed method. Experimental results show that it is books’ authors, genres and themes that consumers consider of mostly when they choose a book commonly. Furthermore, when it comes to a novel, the feature character is also one of the focuses while other features are not as important as them. This research informs a way for enterprises to choose emphasis in an online promotion, which can be very helpful for recommending products to target customers in a proper way.