2012 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (2010)
Aug. 9, 2010 to Aug. 11, 2010
Recently, a huge wave of social media has generated significant impact in people's perceptions about technological domains. They are captured in several blogs/forums, where the themes relate to products of several companies. One of the companies can be interested to track them as resources for customer perceptions and detect user sentiments. The keyword-based approaches for identifying such themes fail to give satisfactory level of accuracy. Here, we address the above problems using statistical text-mining of blog entries. The crux of the analysis lies in mining quantitative information for textual entries. Once the relevant blog entries for the company/ its competitors are filtered out, the theme identification is performed using a highly accurate novel technique termed as 'Best Separators Algorithm'. Logistic regression coupled with dimension reduction technique (singular value decomposition) is used to identify the tonality of those blogs. The final analysis shows significant improvement in terms of accuracy over popular approaches.
Theme identification, Best Separators Algorithm, Sentiment Analysis, Logistic Regression, Singular Value Decomposition
Jayanta Kumar Pal, Abhisek Saha, "Identifying Themes in Social Media and Detecting Sentiments", 2012 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, vol. 00, no. , pp. 452-457, 2010, doi:10.1109/ASONAM.2010.25