AG Angewandte Computerlinguistik
(Applied Computational Linguistics Lab)
[ What? ] -- [ Projects ] -- [ ⇒Demos ] -- [ ⇒People ] -- [ ⇒For Students ] -- [ Contact ]
|
I. Research Statement |
||
|
The research focus of the Applied Computational Linguistics Lab (ALCALI), headed by Prof. Manfred Stede (other members see here), is on theoretical and practical aspects of discourse processing, which includes both text and dialogue. In general we are friends of 'traditional' symbolic---even knowledge-based---methods, but we seek to combine their strengths with the robustness advantages of statistical approaches in various ways. Applications we are currently working on include text summarization (of single and multiple documents), ontology-based information extraction from medical texts, and robust dialog systems for information access. On the theoretical side, we are interested in the structure(s) of discourse and dialogue. As exemplified by our data collection efforts with the 'Potsdam Commentary Corpus', we see multi-level annotation (including the related database maintenance and retrieval issues), as an important tool for discovering regularities and structural properties of discourse. For example, we are interested in modelling the relationship between sentence-internal information structure and the text, or the relation between so-called 'rhetorical structure' and the subjective argument that is brought forward in a commentary. For dialogue, our interests include the mechanisms of clarification in the case of misunderstandings. |
||
II. Projects |
||
|
II.a Summarisation In the SUMMaR project, we develop a prototypical text summarization system. Our goal is to produce high quality summaries while maintaining a high level of robustness. To this end, we combine symoblic, linguistic methods with statistical approaches. In contrast to most current systems, summaries will be partially produced by a text generator ("abstracted") rather than taken over literally ("extracted") from the original source. SUMMaR is part of the BMBF project PINK ("Plattform fuer INtelligente Kollaborationsportale"), a consortium of companies and universities from Berlin-Brandenburg, funded in the framework "Innovative regionale Wachstumskerne". Funding: Bundesministerium für Bildung und Forschung (German Federal Ministry of Education and Research). Members: Dr. S. Dipper, H. Bieler. Project Page, Publications: here. II.b A Linguistic Database for Information Structure The project (full title: ``A linguistic database for Information Structure: Annotation and Retrieval'', part of SFB 632 ``Information Structure'') develops the technical infrastructure for managing the data collected within the large project on Information Structure (SFB 632) (a collaboration between the University of Potsdam and the Humboldt Universität Berlin). This includes developing formats for storing and retrieving all kinds of heterogenous informations, as well as cooperating on developing annotation standards for information structure. Funding: Deutsche Forschungsgemeinschaft (DFG, German Research Council) Members: Dr. Dipper, M. Goetze, C. Chiarcos, J. Ritz Project Page, Publications: here. II.c DEAWU: Dealing with uncertainty in dialogue understanding This EU-funded project aims to improve the way spoken dialogue systems deal with understanding problems, by enabling such systems to produce and understand meaningful and contextually appropriate clarification requests (e.g. ``New York City?'' instead of ``I'm sorry, can you repeat?''). Funding: European Union, Framework Program 6 (FP6) Marie Curie Action "Transfer of Knowledge" Members: Prof. Stede, Dr. Schlangen, Dr. Corradini, Dr. R. Fernandez. Project Page, Publications: here II.d Other Projects We are also running several smaller projects. As one example, we are building a dialogue system tailored for the dialogue sub-genre ''Information-Seeking Chat''. (Project Page: tba.) We are also maintaining and extending a corpus of newspaper commentaries (the Potsdam Commentary Corpus, PCC), annotated on multiple levels, for rhetorical structure, referential links, etc. (Project Page: here.) You can also find a list of current Diploma and PhD theses here. |
||
III. Finished Projects |
||
|
III.a A Semantic Web for Pathology 9/2003--9/2006: In a collaboration with the ``AG Netzbasierte Informationssysteme'' (Net-Based Information Systems Lab) of the Freie Universität Berlin and the Institute for Pathologie of the Charite University Clinic of the Humboldt University Berlin, we developed a system to extract information from medical pathology reports and represent it in OWL -- creating a Semantic Web for Pathology. Funding: Deutsche Forschungsgemeinschaft (DFG, German Research Council) Members: Prof. Stede, Dr. Schlangen. Project Page, Publications: here. |
||
Contact |
|||||
|
|||||