Application of Sentiment Analysis of Web documents for effective personalized Web Search

Resumo

Sentiment analysis is the processing of unstructured data with the primary goal of extracting  people ’ s  views and identifying their moods. In this paper sentiment  of  web documents is used to infer the user intent for effective web search .    The sentiment of the   web document is used  in the recommendation of web documents as per the user’s information need  for better personalized web search. An architecture is designed for web search personalization based on sentiment  analysis of the web documents features. The  web document  are preprocessed to word vector and are associated with sentiment score using Vedar. These labeled web document feature set is further processed using CNN ensemble for feature selection. The resulting labeled web document feature vector are trained using various machine learning algorithms. Multinomial logistic regression displays the high sentiment classification accuracy of web documents. These web documents with positive sentiment score are captured in user s profiling in order to capture user’s  information intent for customized web document recommendations and better search experience.  The comparative analysis of the proposed system  on the dataset of  web documents labeled with sentiment score confirms that use of sentiment score of the web documents in user profiling infer the user’s intent effectively for effective personalized  web search.