Abstract

The comprehensive supply of information from different points of view, e.g., from the thousands of news articles published online every day, is a tremendous advantage of the digital era. However, the immense amount of news material poses a significant challenge to interested readers: It is hardly possible to fully digest this wealth of information, so that the need for systems supporting intelligent news consumption arises. This chapter describes an approach to automatically mining opinions from topically related news article clusters. We focus our work on the extraction of quotations from German news articles and on analyzing the quotations according to the sentiments they express. Our approach is realized as a news aggregation system capable of handling real-world news streams. We describe the architecture and interface of our news aggregator, and present a rule-based method for quotation extraction as well as our supervised approach to sentiment analysis. We evaluate the implemented models on two human-annotated datasets, which can be made available upon request.

@incollection{
year={2015},
isbn={978-3-319-14177-0},
booktitle={Smart Information Systems},
series={Advances in Computer Vision and Pattern Recognition},
editor={Hopfgartner, Frank},
doi={10.1007/978-3-319-14178-7_1},
title={Intelligent News Aggregator for German with Sentiment Analysis},
url={http://dx.doi.org/10.1007/978-3-319-14178-7_1},
publisher={Springer International Publishing},
author={Ploch, Danuta},
pages={5-46},
language={English}
}