A Flexible Parser for a Linguistic Development Environment

Gregor Erbach

In: Otthein Herzog, C.-R. Rollinger (editor). Textunderstanding in LILOG. Integrating Computational Linguistics and Artificial Intelligence. Pages 74-87 Lecture Notes in Artificial Intelligence (LNAI) Springer Berlin 1991.


We describe the parser of LEU/2, the Linguistic Experimentation Environment of the LILOG project. The parser is designed to support and encourage experimentation with different grammars, different styles of writing grammar, and with different parsing strategies. Unlike the parser of the first LILOG prototype, which was designed specifically for Categorial Unification Grammars (Uszkoreit 1986), the present parser places hardly any restrictions on the format of the grammar, and also supports rules in ID/LP-format (Gazdar et al. 1985) under two interpretations of LP statements. Empty categoriescan be processed. The parser includes a mechanism for processing unknown words, and for determining sentence boundaries in continuous text. The parser is a bottom-up chart parser for grammars encoded in STUF (Bouma, König, Uszkoreit 1988). Although the emphasis of the parser is on experimentation with different grammars rather than efficient analysis of texts, we have tried to make the parser as efficient as possible to make it a powerful experimentation tool for the grammar designer. In order to implement parsing strategies, the parsing process is decomposed into individual parsing tasks, which are placed onto an agenda (Kay 1980) according to their priority. This paper consists of two parts. In the first part, we discuss various choices in parser design, and motivate our design decisions by the use of the parser in a linguistic development environment. In the second part, the parser is described in more detail. Some familiarity with chart parsing is assumed

Weitere Links

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz