START Conference Manager    

Towards an Analysis of Opinions in News Editorials: How positive was the year?(project abstract)

Bal Krishna Bal

Eighth International Conference on Computational Semantics (IWCS-8 2009)
Tilburg University, Netherlands, January 7-9, 2009


Summary

Editorials represent opinion articles written by the publisher, editors or columnists of newspapers. From this perspective, editorials are ideal sources for outlining views on events and analyzing how they are perceived, e.g. whether they are positive or negative or the kind of feelings or analysis they involve (uncertainty, worries, etc.). The proposed work aims on identifying the linguistic criteria and creating a model for the aforementioned purpose, which makes use of an adequate set of semantic tags that we have defined to annotate texts from which a detailed analysis can be carried out. Currently, the work is in its preliminary stage, primarily focused on defining the tags required for the semantic-pragmatic analysis. In parallel with the definition of tags and text annotation, we explore ways to construct synthesis of opinions on a given event from various editorials. One of the challenges is to organize the positive and the negative views, and the associated arguments and their strength. We noted that opinions in editorials are not so apparent, which makes synthesis construction a challenging task. As a future work, we plan to develop a computational model which would suggest methods to automate the process of analyzing and synthesizing opinions from editorials. The manually annotated texts and collected editorials would serve as training data and test data respectively for validating the proposed computational model.


START Conference Manager (V2.56.8 - Rev. 414)