University of Twente Student Theses


Matching article sentiment in abstractive summarization of news articles

Regnerus, B. (2019) Matching article sentiment in abstractive summarization of news articles.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Enormous amounts of articles and other texts are published online. In this vast sea of articles, quick access to the most important points is critical to aid people in making decisions on which articles they should read. One of the methods for this is the technique called summarization. Manual summarization is time consuming and no longer always feasible due to the vast amount of text published online. Automatic text summarization has consequently become popular. From the publishers' side however the goal of text summarization can also be to attract readers to the original article. Attracting a readers' attention in text is widely studied by advertisers, however is not used in automatic text summarization. The aim of the final graduation project therefore is focussed on creating and evaluating an automatic abstractive text summarization system which is able to attract a reader's attention. In order to attract a reader's attention there was tried to match the emotion of the source article. There was investigated if it is possible to generate a machine learning model that is able to match the emotion of the source article in the generated summary. Furthermore with a human evaluation it was evaluated if there is a difference in how different summaries are perceived. Significant differences have been observed in how a reader perceives summaries generated in a neural way and summaries where the emotion of the summaries has been changed. This suggests that there is a difference in emotion as perceived by human evaluators between these two. However it can not be concluded based on these results, that the summaries that match the sentiment of the source article therefore attract the reader's attention more than the baseline summaries.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Interaction Technology MSc (60030)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page