ATRIO-USC: Text analysis and retrieval of information for opinion mining: information retrieval and machine learning

A growing number of consumers consult the web before buying any product. Customers give more importance to user-generated contents than manufacturers websites when they make their buying decision. For example, blogs where very different people express their opinions, reviews and experiences about a product and establish comparisons with other similar. In this context, the opinion mining tools stand as a valuable one for consumers because allow them to aggregate and summarize opinions, and enable manufacturers to know the valuations of their users. However, mining tools have two major drawbacks:

  1. apply a very superficial knowledge of the texts, leading to wrong analysis and / or confused;
  2. their ability to manage texts in Spanish is very limited or nonexistent.

This project aims to combine the experience of the teams participating in information retrieval, natural language processing and machine learning to develop new techniques to recover and characterize opinions; extract relationships and comparisons; combine all the information obtained; and determine the sense of the opinions expressed in Spanish as their primary language development. Thus, it promotes the Hispanic community to the area of opinion mining and helps to reduce the technology gap between English and other languages.

Objectives

  • Define the technologies needed to apply opinion mining techniques to expressed opinions in Spanish.
  • Create and compile appropriate support resources for opinion mining tasks on Spanish texts.
  • Design and implement effective methods to locate, obtain and review information based on opinions in Spanish.
  • Design, implement and evaluate effective techniques for information retrieval to obtain fragments of views on a particular subject.
  • Design, implement and evaluate methods for estimating the subjectivity of texts and their orientation and present summaries of views to users.
  • Develop a web application that incorporates the methods, resources and techniques described above, and offers opinions on thematic consultations or specified by the user.