Incremental Specification in Context

The SFB 732 terminated on 30th June, 2018. The websites are no longer updated.

We are very sad to announce that Professor Dr. Grzegorz Dogil, Chair of Experimental Phonetics, died on Tuesday, 26.12.2017, after a long illness.

The Collaborative Research Center (CRC) 732 Incremental Specification in Context was funded from 2006-2018 (Phase 1: 2006 -2010, Phase 2: 2010-2014, Phase 3: 2014-2018) by the German National Science Foundation (DFG). In its – in its last phase - 16 research projects, the infrastructure project and the integrated graduate school it brought together more than 40 researchers from the Institute of Natural Language Processing and the Institute of Linguistics at the University of Stuttgart.

In collaborations between the various subfields of Linguistics and Computational Linguistics, the Collaborative Research Center (CRC) 732 has over a total of 12 years studied a property of linguistic expressions – and the elements that they are composed of – which is observable across all levels of linguistic description: most elements are ambiguous or underspecified when viewed in isolation, but when elements are combined to form larger complexes, most of the ambiguities get resolved. Speech sounds that are compatible with a) various phonemes (e.g. unstressed vowel sounds in fast speech) and/or b) different prosodic categories receive a specific interpretation in a given utterance; syncretic morphological forms (like German sie =she/her/they/them) are disambiguated given the syntax of the sentence; the reading of deverbal nouns like construction, which can refer to an event or its result, is narrowed down through modifiers such as ongoing; etc. At each of the relevant levels of description, it is the elements’ context that drives the disambiguation decision, and the more information becomes available, the narrower is the choice of targets. So what we observe is incremental specification in context. Any account of an aspect of language(s) and language processing must include mechanisms for describing this key ingredient to efficient communicative exchanges, but fully understanding how all relevant levels interact has remained a major challenge in the study of language: Is it specification/disambiguation at one level that triggers further specification decisions at another or vice versa? Or should one assume simultaneous specification decisions? By pursuing these questions in depth for a broad range of linguistic elements, the CRC 732 has significantly enhanced our systematic understanding of language and of speech and language processing. Research contributions range from theoretical advances in various different frameworks over improvements of language-technological models and methodologies to data resources such as speech and text corpora and computational analysis tools.

Starting with the second phase of funding (July 2010), the SFB 732 is complemented by an integrated graduate school (MGK).

The goal of the integrated graduate school is to provide the PhD students in the SFB with a structured qualification program for academic training and a forum for mutual exchange and feedback. To this end, a teaching program aims at broadening the background and improving the professional opportunities of the PhD students, to ensure further qualification of the graduate school members. Additionally, graduate school members meet regularly to present and discuss their postgraduate work in an equal-status group and to encourage collaborative and interdisciplinary research.

Short-term one-year research grants for PhD students serve as a special tool for mutual exchange and collaborative work. The grants are closely linked to individual projects of the SFB.

Please check out the MGK site for more information.

• Institute of Linguistics
• IMS (Institute for Natural Language Processing)
• LSS (Lehrstuhl für Systemtheorie und Signalverarbeitung)
• FSK (Forschungsverbund Sprachwissenschaft und Kognition)
• DFG (Deutsche Forschungsgemeinschaft)
• Lingenio (for information on cooperations with the IMS and the SFB click on the respective links)