Today E-commerce websites provide customers with the needed product information by giving various kinds of services to choose from. One such service is to allow the customer to read the end user online reviews. Online reviews contain features which are useful for the analysis in opinion mining. Converting these unstructured reviews into useful information require extracting the product features from them. Natural Language Processing (NLP) based technique extracts various kinds of product features including the low frequency features. Topic Modeling based approach also identifies specific product features from the online reviews. The effective number of product features is made available to the customer when these two approaches are combined. This results in the expanded product feature set so that the customer makes wise decisions without having to compromise on the feature set.

Online product reviews provide data about the user's perspective on the features that were experienced by them. Product features and corresponding opinions form a major part in analyzing the online product reviews. Extracting features from a huge number of reviews is classified into three major categories such as utilizing language rules, sequence labeling as well as the topic modeling. Latent Dirichlet Allocation (LDA) is one such topic model which clusters the document words into unsupervised learned topics using Dirichlet priors. The words so clustered are the features and opinion words in the product reviews domain. To identify appropriate product features from these clusters a hierarchical, domain independent Feature Ontology Tree (FOT) is applied to LDA clusters. The opinion bearing words of obtained product features are identified by utilizing the document indicators available from topic matrix of LDA. These indicators are useful to backtrack to the corresponding online review in which the product feature is present. The polarity of the opinion bearing word is calculated with the help of SentiWordNet. This improves the accuracy of the features using extracted LDA topic clusters and machine interpretation of polarity of opinion word is satisfactory.