Open domain context-based targeted sentiment analysis system

Massive amount of opinions is available nowadays on the internet thought social networks. Availability of such wealth of opinions in the social media motivated researchers to develop automated systems for opinion mining, also known as sentiment analysis. A very recent direction currently addresses the problem of detecting the targets, i.e., topics, in the micro-blogs (e.g., tweets) and identifying the sentiment polarity toward them. This research direction is referred to as open domain sentiment classification. In this work, we developed an approach for identifying the context of a given set of micro-blogs and analyzing their sentiments within that context. We propose a new context-based analysis system for open domain targeted sentiment analysis. Based on our knowledge, the proposed system differs from status quo followed in developing systems for open domain targeted sentiment analysis. Existing systems detect opinions in each micro-blog individually. While, our proposed approach helps in developing a context-based analysis system among a set of micro-blogs. The proposed system detects context patterns among a set of micro-blogs by detecting targets and identifying sentiment polarities towards categorized topics that describe the detected targets. Experimental results shown in this paper illustrate performance of applying the proposed system to Arabic and English datasets. Some comparisons between the proposed system and other existing system are included in this work as well.