Prof. Dr. Manfred Stede (Universität Potsdam)
12. Juni 2014
Sentiment analysis has become one of the most popular applications of computational linguistics, with one reason being its enormous potential for commercial application: For example, the automatic mining of product reviews is of interest to many vendors of goods and services on the web. The technical approaches can be broadly distinguished into the categories of (i) corpus-based machine learning involving no linguistic knowledge, and (ii) lexicon-based methods that employ a certain amount of linguistic analysis. Both approaches have been used in my research group. In the talk, I will explain our implementations and discuss their relative merits, thereby taking a somewhat broader perspective on the role of "subjectivity" in text mining.
Manfred Stede is a professor of Applied Computational Linguistics in the Linguistics Department of Potsdam University. He obtained his Ph.D. from the University of Toronto in 1996 with a thesis on multilingual text generation. Prior to assuming the position in Potsdam in 2001, he worked on the 'Verbmobil' machine translation project at TU Berlin from 1995 to 2000, and for one year joined Semantic Edge GmbH (Berlin) to work on automatic dialog systems. His recent projects in Potsdam on the applied side dealt with linguistic databases, text summarization, dialogue systems, and sentiment analysis; on the theoretical side he is interested in models of text coherence and text structure.