Colloquium: Data2Text: The Automatic Generation of Dialogue Utterances, ranslations and Reports

29 May 2017

Computers can store large amounts of data that are hard to oversee and to keep track of. This talk explains some of the ways how computers can produce meaningful and purposeful text from non-linguistic data. We look at three different areas for which new text is needed. (Spoken) dialogues have become rather common on smartphones or with in-car applications. How does the system know to express itself in such a way that it meets the user's point of interest? Second, if we want the computer to take a text and rewrite it in some other language, how can does the system ensure the result entails the same meaning? And finally, in the many situations where we simply want a linguistic overview of a given selection of data, how does the system produce a well-structured, coherent and fluent text? With a focus on the latter part, we will investigate different ways of text generation and show their promises and limitations.

Monday, 29 May 2017, Uni Hamburg, Hörsaal ESA-K, Edmund-Siemers Allee

Prof. Dr. Stefan Busemann,  Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI)