Feature-based customer review summarization

Abstract

To systematically monitor the online customer satisfaction means to deal with a large amount of non-structured data and with many Natural Language Processing challenges. The purpose of the present research is to automatically identify the benefits and the drawbacks expressed by internet users in Italian customer reviews in free text format. The work is grounded on Italian lexical and grammatical resources that, together, are able to investigate the semantic relation between the product features and the opinions expressed on them. On the base of these resources we built DOXA, a linguistic-based application that gives a feedback of statistics about the positive or negative nature of the opinions and about the semantic categories of the features.