PROJECT TITLE :

Collaboratively Training Sentiment Classifiers for Multiple Domains - 2017

ABSTRACT:

We have a tendency to propose a collaborative multi-domain sentiment classification approach to train sentiment classifiers for multiple domains simultaneously. In our approach, the sentiment information in different domains is shared to train more accurate and strong sentiment classifiers for every domain when labeled data is scarce. Specifically, we have a tendency to decompose the sentiment classifier of every domain into 2 parts, a international one and a website-specific one. The world model can capture the general sentiment information and is shared by various domains. The domain-specific model will capture the particular sentiment expressions in each domain. In addition, we extract domain-specific sentiment data from both labeled and unlabeled samples in each domain and use it to boost the learning of domain-specific sentiment classifiers. Besides, we incorporate the similarities between domains into our approach as regularization over the domain-specific sentiment classifiers to encourage the sharing of sentiment info between similar domains. 2 sorts of domain similarity measures are explored, one primarily based on textual content and the other one based on sentiment expressions. Moreover, we introduce two efficient algorithms to solve the model of our approach. Experimental results on benchmark datasets show that our approach will effectively improve the performance of multi-domain sentiment classification and considerably outperform baseline strategies.