English Resource Grammar

The LinGO English Resource Grammar (ERG) is a broad-coverage, linguistically precise HPSG-based grammar of English. It was developed initially using the DFKI PAGE system, but the LKB is now the primary grammar engineering environment. The ERG is semantically grounded in Minimal Recursion Semantics (MRS), which is a form of flat semantic representation capable of supporting underspecification.

The first application of the LinGO ERG was in the Verbmobil spoken language machine translation project. CSLI has been responsible for building the English grammar for the deep-processing component of Verbmobil, which utilizes a semantic transfer approach. The grammar therefore contains good coverage of the constructions most frequently found in the Verbmobil data which concerns meeting scheduling and travel reservations: it can currently produce semantic representations for about 83 per cent of the utterances in a corpus of transcriptions of some 10,000 utterances, which vary in length from one word to more than thirty words. The hand-built lexicon of around 10,000 words (current July, 2003) is somewhat tuned to this domain, augmented more recently to accommodate the vocabulary found in electronic commerce email messages studied by some of our industrial affiliates.

Dan Flickinger (CSLI) is the principal ERG developer. Other individuals who have made major contributions to the grammar are Emily Bender (Professor, University of Washington), Ann Copestake (Reader, University of Cambridge), Rob Malouf (Professor, San Diego State University) and Stephan Oepen (Professor, University of Oslo). Several former graduate students (Brady Clark, Judith Tonhauser, Kathryn Campbell-Kibler, Martina Faller, Ash Asudeh, Susanne Riehemann) and visiting graduate students (Jesse Tseng, University of Edinburgh; Ken Bame, Ohio State University; Judith Eckle-Kohler, University of Stuttgart) also did detailed work, including building the lexicon, developing test suites, isolating phenomena found in corpora and developing analyses in the HPSG formalism. Jeff Smith (Professor, San Jose State University) has also spent time at CSLI developing various aspects of the grammar. In addition to this direct implementation work, weekly technical project meetings have provided an important forum for critique of specific analyses, particularly from Ivan Sag (Professor, Stanford) and Tom Wasow (Professor, Stanford).

References

Dan Flickinger (2002) On building a more efficient grammar by exploiting types. In Stephan Oepen, Dan Flickinger, Jun'ichi Tsujii and Hans Uszkoreit (eds.) Collaborative Language Engineering, Stanford: CSLI Publications, pp. 1-17.

Ann Copestake, Dan Flickinger, Ivan A. Sag and Carl J. Pollard (1999) Minimal Recursion Semantics: An Introduction.

Ann Copestake and Dan Flickinger (2000) An open-source grammar development environment and broad-coverage English grammar using HPSG In Proceedings of the Second conference on Language Resources and Evaluation (LREC-2000), Athens, Greece.