Review based Feature Matrix for Predicting ratings in Recommender System

Abstract

Abstract-Recommender systems are acquiring extensive popularity and have become essential component of on-line business handling tools because of their capability of providing personalized guidance in selecting products and services. Collaborative Filtering that brings in the popularity aspect of the item amongst the user base, heavily depends on the ratings provided by the users, while Contentbased Filtering that brings out the item's features matching the user's taste, requires content information as also user's preference information. Service providers usually invite users to share their experience about the use of service in the form of reviews and ratings. Reviews which are verbose contain a rich source of information about the service's features as also user's preferences while ratings are usually sparse due to user's reluctance to quantify. A feature matrix generated by processing review information using semantic similarity based on synsets can be used along with sparse ratings to generate the complete predicted ratings matrix. The paper presents a modified matrix factorization approach for recommendations using the review based feature matrix Keyword-Recommender system, synsets, predictive model. I. INTRODUCTION Recommender systems(RS) have become common place in the digital market arena, by providing the on line user of products and services, a personalized selection guidelines. Generally technology used by RS is categorized into two groups as Content-based and Collaborative [1]. Content based systems are based on user profile and item description, giving importance to various features of items that may interest the user or match his taste.It recommends the item that is having the features that user has liked in the previously consumed items [2].Collaborative Filtering harnesses on popularity element or standing of the item across theuser base and works by collecting user feedback in the form of ratings of items. The challenges faced by collaborative filtering are i. The sparsity of rating matrix due to reluctance on the part of user to rate all the items. ii.The unavailability of data for new user or new item which is also known as cold start problem. iii. Hijacking of the RS by pushing own product rating to the higher value and lowering the competitor's rating leading to reduction in the quality of recommendations. Matrix factorization is widely used, model based approach that estimates ratings. It is a class of latent factor models, where user and items are represented as unknown feature vectors along latent dimensions. The feature vectors are learned using known ratings and learnt features are then used to predict unknown ratings. In the presence of sparse ratings, reviews in textual form present a rich source of feature information. The reviews are converted into the Feature Matrix with group of terms based on synsets as a feature, utilizing Semantic relationship existing between the terms in the form of Synonyms and meronyms. In this paper Matrix factorization approach uses the above generated fe...

Marathi Text Analysis using Unsupervised Learning and Word Cloud

“…Various data mining techniques like clustering, classification can be applied once the data is in a structured format. Document term matrix (DTM) allows to convert text files into table form, where rows are represented by documents and terms are placed as columns [45][46][47]. But DTM causes dimension curse because all the terms present in the corpus are considered while constructing DTM.…”