Sentiment Neural Network(SNN)-for Knowledge based Recommender System

Abstract

Recommendation systems play a crucial role in e-commerce by recommending products that suit for the consumers also providing exact information to users. For the past few decades, researchers used many machine learning techniques on recommender system. In recommender system ML based algorithms providing a better detection of user preferences, item features and users-items. In this way deep learning play an increasingly vital role in review recommender systems, since they used a bunch of discrete values for review. However, a problem arises regarding that feedbacks. Discrete values are those which are hard to describe user’s interests. This problem makes it impossible to specifically model user’s choice for recommendation. Purpose of this paper is to introduce a new novel SNN (Sentiment Neural Network) framework for effective recommendation system. SNN frameworks consider two phases. In first phase of SNN framework, this work introduces NLP techniques for converting unstructured data into structured data. This technique includes pre-processing data, feature extraction, word scoring, polarity classification and sentiment analysis. The second phase of SNN framework used for verifying the polarity classes with the real world examples. SNN structures are not only among the best models with reference to prediction accuracy, they also consider the weighting factor on classifier to reduce the training time. A novel Sentiment Neural Network along with knowledge recommender system is suggested for review features extraction, text classification and analyzing review features in the various domains.