Sentiment Analysis on Micro-Blogs

Abstract

Sentimental analysis has been growing among the researcher and in business due to ongoing importance in many domains such as feedback system, review system, individual’s sentiment etc. Due to availability of large corpus of texts as of now the supervised models are now started to yield good result. But text is not quite straight forward it is needed to cleaned and prepared for further processing. In this research we aim to create a different model so as to develop and predict sentiments on different categories of micro-blogs obtained on social-media. The approach we to tried here is to construct many existing theories and approaches and then apply the proposed model. We used pre-trained GloVe model for word representation which is basically an unsupervised learning algorithm trained on many text corpuses. The Proposed Model mentioned above is build using CNN, Dense Network and unidirectional LSTM and GRU cells of multiple layers. The output obtained is categorical ranged between 0-10. This model is than used as a back-end to evaluate summarized sentiment of posts with sentiments of each micro-blogs from the respective post. The fetching and transferring of data done with help of REST-API. This type of Learning-Model and Fetching technique enhances the accessibility and availability of service.