Automatic opinion lexicon extraction has attracted lots of attention and many methods have thus been proposed. However, most existing methods depend on dictionaries (e.g., WordNet), which confines their applicability. For instance, the dictionary based methods are unable to find domain dependent opinion words, because the entries in a dictionary are usually domain-independent. There also exist corpus-based methods that directly extract opinion lexicons from reviews. However, they heavily rely on sentiment seed words that have limited sentiment information and the context information has not been fully considered. To overcome these problems, this paper presents a word vector and matrix factorization based method for automatically extracting opinion lexicons from reviews of different domains and further identifying the sentiment polarities of the words. Experiments on real datasets demonstrate that the proposed method is effective and performs better than the state-of-the-art methods.
[1]
H. Sebastian Seung,et al.
Learning the parts of objects by non-negative matrix factorization
,
1999,
Nature.
[2]
Li Guo,et al.
CONR: A Novel Method for Sentiment Word Identification
,
2014,
CIKM.
[3]
Jeffrey Dean,et al.
Distributed Representations of Words and Phrases and their Compositionality
,
2013,
NIPS.
[4]
Michael L. Littman,et al.
Measuring praise and criticism: Inference of semantic orientation from association
,
2003,
TOIS.
[5]
Hongliang Yu,et al.
Identifying Sentiment Words Using an Optimization-based Model without Seed Words
,
2013,
ACL.