Institute for Web Science and Technologies · Universität Koblenz - Landau
Institute WeST

Mining Text and Social Media Papers at 4 Top Conferences

[go to overview]

From February 24-28, the Web Search and Data Mining (WSDM) conference took place in New York, USA. WSDM is a premier and highly selective conference covering search and data mining on the web. Here, Christoph Carl Kling from the WeST Institute presented the paper “Detecting Non-Gaussian Geographical Topics in Tagged Photo Collections”. The novel approach allows for an improved detection of geographical topics by exploiting adjacency relations between spatial regions. Existing approaches do not support the detection of geographical topics.

Furthermore the International AAAI Conference in Weblogs and Social Media (ICWSM) is a leading conference blending social sciences and computer science. It takes place in Ann Arbor, Michigan, USA, from June 2-4. This year, Claudia Wagner and Markus Strohmaier from WeST will present research conducted in GESIS - Leibniz Institute for the Social Sciences. The paper „When politicians talk“ shows an empirical study on the conversational practices of German politicians on Twitter before, during and after the German federal elections in 2013. Besides concrete results the authors describe their general socio-computational approach for analysing conversational practices.

Thirdly in April 2014 the prestigious 23rd International World Wide Web conference (WWW) will take place in Seoul, Korea. WWW covers topics of the future direction of the World Wide Web. There, Claudia Wagner and Markus Strohmaier from the WeST Institute will present joint work with Philipp Singer from Graz University of Technology and Bernardo Huberman from HP Labs. The paper “Semantic Stability in Social Tagging Streams” discusses existing methods for measuring semantic stability in social tagging systems. Based on identified shortcomings the authors present a novel and more suitable approach for measuring and comparing the semantic stability in different social stream based systems.

Finally this year the conference of the Association for Computational Linguistics (ACL 2014) will take place in Baltimore, USA, from June 23-25. Rene Pickhardt, Thomas Gottron, Martin Körner, Paul Wagner, Till Speicher and Steffen Staab present their paper “A Generalized Language Model as the Combination of Skipped n-grams and Modified Kneser Ney Smoothing”. The paper describes a generalized Language Model with modified Kneser Ney Smoothing that improves the current state-of-the-art approach with an decrease in perplexity of about 10%. The current state-of-the-art method was not beaten for the last 15 years. This is of particular interest as language models are widely spread in natural language processing due to the popularity of the Noisy channel model. The source code (and in future also the paper) can be found at https://github.com/renepickhardt/generalized-language-modeling-toolkit. The research is based upon the work of two former students Paul Wagner and Till Speicher who successfully participated at the German competition Young Scientists and who were advised by René Picard.


21.03.14

Contact for inquiries: west@uni-koblenz.de