+7 (495) 987 43 74
Join us -              
Рус   |   Eng


Yatsko V.

Doctor of Science, Full Professor, Katanov State University of Khakasia

A system for automatic customers’ opinion mining

which was created on a request from an overseas firm, processes chats, blogs, and forums to provide as output generalized negative and positive scores as well as specific subjectivity terms associated with the product’s name given in user’s query. The system is based on the highly structured ontology and linear grammar. The ontology consists of syntactic and semantic terms ad their subcategories, groups and classes. Semantic terms are those ones expressing subjectivity and having either positive or negative semantic orientation. Syntactic terms do not express subjectivity but can modify the intensity of polarity of semantic terms. The relation between semantic and syntactic terms is defined as a binary relation and groups with symmetric, asymmetric, and inverse relations between syntactic and semantic terms are distinguished. Semantic terms are assigned scores from 1 (–1) to 9 (–9). The paper suggests differentiating between formal and linguistic ontologies. Linear grammar comprises rules that allow mapping ontology terms onto the names of products and generate phrases with semantic and syntactic terms. The overall system’s efficiency measured in terms of recall and precision of information retrieval achieves 86%.