In progress Time-Based Aspect-Level Sentiment Analysis (TALSA)

Reference:
ERIM PhD 2012 ESE LIS RD_FJ_FF

Abstract

With opinions and expressions of sentiment available in abundance on the Web, algorithms that can deal with this kind of subjective information abound. While several approaches exist that perform aspectlevel sentiment analysis, which associates the expressed sentiment to aspects and entities instead of linguistic constructs like sentences or documents, the temporal dimension is still lacking. By incorporating the temporal aspect of sentiment, we can correctly aggregate sentiment expressed at varying points in time and perform trend analysis by looking at how sentiment changes over time. The main goal is to gather sentiment information from the Web and aggregate it into knowledge that is useful for business applications such as product marketing, product comparison, or reputation management of a company, brand, or person. Our hypothesis is that an incremental, online machine learning model is particularly well suited to capture these temporal elements in addition to performing aspect-level sentiment analysis as defined in the current literature.

Keywords

text mining; linguistic processing; machine learning; text analysis; sentiment analysis; aspects; temporal dimension

Time frame

2012 - 2016

Supervisory Team

Rommert Dekker
Professor of Econometrics (Operations Research and Informatics)
  • Promotor
Franciska de Jong
Franciska de Jong
  • Promotor