A rule-based, domain independent approach for opinion and holder identification

摘要

Mining sentiments from text is currently an important problem in information retrieval systems. In this paper we propose a solution for extracting opinions and opinion holders from large texts. Our goal is to achieve a high level of domain independence by implementing a rule-based approach. The results of our system have proven an accuracy which is comparable to that of systems that use a supervised learning approach, which is domain dependent.