Automatic Opinion Analysis Based on SVM and Distance-Weighted Computing

This paper presents an SVM and distance-weighted computing based approach to analyze opinion, which includes two aspects: opinion automatic detection and opinion holder automatic extraction. We established a series related resources to opinion analysis, such as opinion-bearing verb set, orientation indicating adverb set etc. Based on these related resources, we accomplish the detection of opinions via SVM. On the basis of opinion automatic detection, we use opinion-bearing verb as a key indicator and apply distance-weighted computing method to extract the opinion holder (or holders) of the given sentence. With expanding the holders to noun phrases by pattern matching, the results of opinion holder automatic extraction are further improved.