Paper Details

Sentiment classification is a topic-sensitive task, i.e., a classifier trained from one topic will perform worseon another. This is especially a problem for the tweets sentiment analysis. Since the topics in Twitter are very diverse, itis impossible to train a universal classifier for all topics. Moreover, compared to product review, Twitter lacks datalabeling and a rating mechanism to acquire sentiment labels. The extremely sparse text of tweets also brings down theperformance of a sentiment classifier. In this paper, we propose a semi-supervised topic-adaptive sentiment classification(TASC) model, which starts with a classifier, built on common features and mixed labeled data from various topics. Itminimizes the hinge loss to adapt to unlabeled data and features including topic-related sentiment words, authors’sentiments and sentiment connections derived from “@” mentions of tweets, named as topic-adaptive features. Text andnon-text features are extracted and naturally split into two views for co-training. The TASC learning algorithm updatestopic-adaptive features based on the collaborative selection of unlabeled data, which in turn helps to select more reliabletweets to boost the performance. We also design the adapting model along a timeline (TASC-t) for dynamic tweets. Anexperiment on 6 topics from published tweet corpuses demonstrates that TASC outperforms other well-known supervisedand ensemble classifiers. It also beats those semi-supervised learning methods without feature adaption. Meanwhile,TASC-t can also achieve impressive accuracy and F-score. Finally, with timeline visualization of “river” graph, peoplecan intuitively grasp the ups and downs of sentiments’ evolvement, and the intensity by color gradation.

Prof. Dyaneshwar Kudande, Swapnil Desale, Abhishek Kolte, “TASC:Topic-Adaptive Sentiment Classification on Dynamic Tweets”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 4, Issue 3, pp. 497-500, March 2017.