Aspect extraction from reviews using conditional random fields

Abstract

This paper describes an information extraction and content analysis system. The proposed system is based on a conditional random field algorithm and intended to extract aspect terms mentioned in the text. We use a set of morphological features for machine learning. The system is used for automatic extraction of explicit aspects and also to automatic extraction of all aspects (explicit, implicit and sentiment facts), and tested on two domains: restaurants and automobiles.We show that our system can produce quite a high level of precision which means that the system is capable of recognizing aspect terms rather accurately. The system demonstrates that even a small set of features for conditional random field algorithm can perform competitively and shows good results.