PhD Topics

Aspect-Based Sentiment Analysis

Guidelines

Under sentiment analysis (SA), we understand the task of automatic extraction of an opinion from a natural text. The prototypical applications of SA include summarizing users' views from discussions or market trends prediction based on consumers' data.

The aim of the thesis is to experiment with methods refining SA to the level of 'aspects'. Instead of reporting the general sentiment of a given text (e.g. whether a review overall does or does not recommend a product), the thesis will be automatically extracting the set of mentioned aspects of the product together with the expressed valuations (e.g. that the display of a cell phone is outstanding but
the battery is poor).

The primary language of experiments will be Czech, taking advantage of both existing SA techniques designed for English as well as the range of NLP tools available for Czech. The thesis will review existing SA methods and adapt them as needed for Czech. Whenever possible, the proposed methods should be language independent, but we expect that the best performance can be reached only when employing language-specific knowledge. This also holds for domain dependency.

An empirical evaluation using standard quality measures is an inherent part of the thesis. We can benefit from extensive knowledge developed at ÚFAL about SA related tasks and use and further develop already existing data such as Czech Subjectivity Lexicon or sentiment extension of PDT data.

References

SemEval 2014. Proceedings with the results of the competition - Task 4: Aspect Based Sentiment Analysis. In preparation.

Liu, Bing. Sentiment analysis and opinion mining. Synthesis Lectures on Human Language Technologies 5.1 (2012): 1-167.

Blitzer, John, Mark Dredze, and Fernando Pereira. Biographies, bollywood, boom-boxes and blenders: Domain adaptation for sentiment classification. Proc. of ACL. Vol. 7. 2007.

Brody, Samuel, and Noemie Elhadad. An unsupervised aspect-sentiment model for online reviews. Proc. of HLT-NAACL. 2010.

Pang, Bo, and Lillian Lee. Opinion mining and sentiment analysis. Foundations and trends in information retrieval 2.1-2 (2008): 1-135.

Snyder, Benjamin, and Regina Barzilay. Multiple Aspect Ranking Using the Good Grief Algorithm. Proc HLT-NAACL. 2007.

Zhai, Zhongwu, et al. Clustering product features for opinion mining. Proceedings of the fourth ACM international conference on Web search and data mining. ACM, 2011.