An LDA-based Approach for Product Attribute Identification from Online Customer Reviews

Abstract

This paper proposes an Latent Dirichlet Allocation (LDA)-based approach to identify product attributes from online customer reviews. Identifying product attributes from the customers’ perspectives is essential to analyze satisfaction, importance, and Kano category of each product attribute for product design. The previous works overlooked the importance of keyword extraction and filtering in keyword preprocessing. The proposed approach provides an automated method to select product-feature words by improving manual works in keyword preprocessing of LDA. This research can consider noun phrases as product-feature words and group product-feature words that are frequently mentioned together by customers into the same product attribute better than the previous approach based on the word similarity. The case study of android smartphones shows that the proposed approach can identify product attributes better than the previous approaches.