Information Retrieval and Sentiment Analysis for Scientific Literature

Master’s defense by Youhui Shi


Sentiment analysis for information retrieval in scientific literature is completely unexplored.

An example of sentiment analysis for information retrieval in scientific literature is being able to search for positive, disappointing or contradictory findings about a scientific method.

In this thesis, we investigate the effects and potential usefulness of sentiment analysis upon the effectiveness of information retrieval for scientific literature. We first investigate whether and to what extent sentiment exists in scientific literature. We then propose an information retrieval model that ranks documents with respect to queries, not only according to their topical relevance, but also according to the expressions of sentiment in the document. We present different variants of this model, and implement it in a working information retrieval application. Evaluation on a collection of scientific literature indicates that the inclusion of sentiment analysis in information retrieval for scientific literature can improve retrieval performance compared to state of the art baselines.

The defense is public and will be held in English.

Place: DIKU-SC Mødelokale 24-5-62

Supervisor: Christina Lioma, DIKU

External examiner: Assoc. Professor Troels Andreasen, Roskilde University