University Fachrichtung Up

Current Event: "Reusability of Grammatical Resources" Workshop (April 6-7, 1995)

Reusable Grammatical Resources (RGR)

Developing large grammars for language-processing systems is immensely laborious -- the development of the most successful computational grammars has taken at least a decade. In spite of the large costs involved, such grammars do not survive the software systems for which they were produced. The encoded linguistic knowledge thus cannot be systematically translated into new description languages. What has been missing is primarily a suitable mathematical formalization of the data structures and operations used. Through the convergence of description formalisms in computational linguistics during the last ten years, there is now a basis for the development of data types that is applicable to most formalisms.

In the project, polytheoretic abstract data types for linguistic description formalisms will be developed in cooperation with working groups from the universities of Utrecht, Tilburg and Edinburgh. The basis of the concrete encoding is the ALEP formalism, which was developed in an earlier EC project in which computational linguists from our institute participated. The results of the project are an overview of datatypes used in theoretical and computational linguistics, a library of linguistic datatypes implemented in the ALEP formalism, and extensions of the ALEP formalism with external constraint solvers for datatypes like sets or tree descriptions.

The following datatypes and operations were selected because they enjoy widespread use in linguistic descriptions, and because they offer additional expressive power over current formalisms.

Researchers in Saarbrücken have worked on the datatypes used in LFG and HPSG, and on the implementation of set constraints, linear precedence constraints and tree descriptions. The project is funded by the European Community through the programme "Linguistic Research and Engineering"

Official RGR Project Description