Date: Wed, 20 Nov 1996 19:08:49 GMT Server: Apache/1.1.1 Content-type: text/html Content-length: 4257 Last-modified: Tue, 12 Nov 1996 16:08:42 GMT
This is the homepage of the Research Lab for Linguistics and Computation at the Computer Science Department of Brandeis University , Waltham MA, USA.
RLLC is involved in research on the design and development of lexically-based semantic indexing and content abstraction techniques for documents. These techniques employ robust natural language processing of text to generate multi-purpose structures which we call lexical webs Lexical webs are normalized structured representations of the semantic content of a text or group of texts. An important application of lexical webs is their potential role in accelerating the placement of information on the web in a highly indexed format, which facilitates navigation and rapid information access (see TexTract demo).
This work emerges directly from previous research carried out under the NSF-ARPA Human Language Technologies program. Our goal in this research has been the development of an intelligent and trainable lexical database for a substantial fragment of English, called the core lexical engine Central to this work is an integration of lexical semantic techniques with corpus-based approaches to lexical acquisition. By characterizing the statistical properties of texts which have been annotated with semantic tags, one can infer semantic relations in unmarked texts that are richer than mere collocational associations.
Future research is directed towards techniques for automatically enriching the model of web-based documents by making use of and extending these technologies. It takes advantage of the fact that the generative lexicon framework has developed partly in the context of work within the Tipster and MUC efforts; namely, providing statistically-sensitive semantic techniques for lexically-based document enrichment in the service of information extraction tasks.