Siti Zakaria MSc Dissertation 2014/15

Tell me what is important, and only what is important

Supervised by L.Specia

Abstract

Knowing the opinion of costumers about a given product or service is essential for companies that aim to deliver or keep high quality products and services in order to survive competition, as well as for potential new customers who are interested in a specific product or service. Nowadays it is a common practice for customers to review products or services they buy on internet-based channels. Examples include reviews on social networks such as Twitter or e-commerce websites such as Amazon.

For popular products such as iPhone 6, there can be hundreds or even thousands of reviews containing very useful information for other customers and the company selling the product or service. However, the content of many of these reviews is at least partially irrelevant. For example, people expressing their opinion on other products or on related aspects such as the product delivery. Also, many reviews contain redundant information. More important, it is often infeasible for the potential customers and companies to read each single review. Therefore, techniques to help the automatic identification and extraction of relevant information from a collection of reviews are needed. The task involves investigating and experimenting with different Natural Language Processing (NLP) techniques in order to identify important sentences in customer reviews and extract information from them.