A Neural Network Based Approach for Text-Level Sentiment Analysis Using Sentiment Lexicons

There have been many discussions on forums, e-commerce sites, sites for reviewing products, social media which helps in exchanging opinions, thoughts through free expression of users. Internet as well as web 2.0 is overflowing with the data generated by users which provides a good source for various sentiments, reviews, and evaluations. Opinion mining more popularly known as sentiment analysis classifies the text document based on a positive or negative sentiment that it holds. This is an open research domain and this particular research paper puts forth a model called Artificial Neural Network Based Model i.e., ANNBM. The model is trained and tested through Information Gain as well as three other popular lexicons to extract the sentiments. It’s a new approach that best utilizes the ANNBM model and the subjectivity knowledge which is available in sentiment lexicons. Experiments were conducted on the mobile phone review as well as car review to derive that this approach was successful in finding best output for sentiment-based classification of text and simultaneously reduces dimensionality.