Research on Product Knowledge database Construction Using Large-scale Online Reviews
Abstract：The rapid develop of Web 3.0, knowledge database construction, serves as an important technology for supporting intelligence systems, receives much attention. Constructing domain-specific knowledge database could help to achieve the goal of setting up vertical search engines and making machines to be able to learn and inference. This research aims to study the problem of product knowledge database construction. The schema of product knowledgebase is quite complicated and requires real-time update, which makes conventional knowledgebase construction method infeasible to create product knowledge database. At the same time, contents of most current knowledge database are certain and object knowledge. However, for products, customers’ opinions are valuable information, which are indispensable for a product knowledge database. Therefore, this research aims to build product knowledge-based with customers’ opinions, which uses large-scale online reviews as the data source. The key problems include: (i) Using large-scale online reviews to automatically build up product feature schema; (ii) Exploiting probabilistic fine-grained opinion mining methods with deep semantic comprehension; (iii) Filtering out fake reviews based on latent factor analysis; (iv) Integrating the constructed product knowledge database to our vertical search engine to test and verify its effectiveness.
Keywords: knowledge database construction; opinion mining; online reviews
Contact:
XU Liheng
E-mail: lhxu@nlpr.ia.ac.cn