Title :
A grammatical dependency improved CRF learning approach for integrated product extraction
Author :
Hui Song ; Yan Yan ; Xiaoqiang Liu
Author_Institution :
Coll. of Comput. Sci. & Technol., Donghua Univ., Shanghai, China
Abstract :
The aspect-based opinion mining aims to provide finegrained product feature analysis from the product reviews. Nowadays, newer works based on probabilistic models have achieved satisfactory result on product entities extraction, but these works didn´t take the multi-words feature expression problems into consideration which lead to inaccurate match between the feature and the sentimental orientation. In this paper, we propose an approach, based CRF learning model, to extract integrated feature expression from Chinese product reviews, and improve it by grammatical dependency generated by Stanford Parser. The experiment results based on actual data, demonstrate that the proposed approach is effective and domain-independent. Also we find introducing grammatical dependency into CRF model can improve the precision and recall in varying degrees.
Keywords :
data mining; grammars; learning (artificial intelligence); reviews; statistical analysis; Chinese product reviews; Stanford parser; aspect-based opinion mining; conditional random fields; finegrained product feature analysis; grammatical dependency improved CRF learning approach; integrated product extraction; multiwords feature expression problems; probabilistic models; sentimental orientation; Aspect-based; Conditional Random Fields (CRFs); Domain-Independent; Feature expression extraction; Grammatical Dependency;
Conference_Titel :
Computer Science and Network Technology (ICCSNT), 2012 2nd International Conference on
Conference_Location :
Changchun
Print_ISBN :
978-1-4673-2963-7
DOI :
10.1109/ICCSNT.2012.6526267