Applying an Experimental MT System to a Realistic Problem
608Authors
nan
Abstract
AbstractThis presentation outlines the implementation of a machine translation system for avalanche warning bulletins in natural language, using a unification-based formalism developed at ISSCO, which will be introduced at the same occasion. Concrete examples taken from this project exemplify a modern approach to ma- chine translation: a rich representation of the semantic content of a sentence, the use of a sin- gle grammar for parsing and generating as well as generation and transfer based exclusively on the semantic representation of a sentence. Simultaneously, the limits of bidirectional trans- fer are being tested.
Keywords
Full Content
[MT Summit III, Washington, DC, USA, 1-4 July 1991] Applying an experimental MT system to a realistic problem Pierrette Bouillon and Katharina Boesefeldt ISSCO, 54 Route des Acacias 1227 Geneva, Switzerland pb@divsun.unige.ch kathy@divsun.unige.ch Abstract This presentation outlines the implementation of a machine translation system for avalanche warning bulletins in natural language, us- ing a unification-based formalism developed at ISSCO, which will be introduced at the same occasion. Concrete examples taken from this project exemplify a modern approach to ma- chine translation: a rich representation of the semantic content of a sentence, the use of a sin- gle grammar for parsing and generating as well as generation and transfer based exclusively on the semantic representation of a sentence. Si- multaneously, the limits of bidirectional trans- fer are being tested. 1 Introduction The aim of this presentation is to show how ELU,1 a uni- fication environment developed for linguistic experimen- tation with parsing, generating and translation, is being used for the elaboration of a concrete machine transla- tion system dealing with a limited semantic domain. In this paper we concentrate on the notion of a for mal representation as the basis for mapping between lan- guages. We will show how a semantic representation con- sisting of a complex feature structure enables an efficient transfer between languages in this limited domain. 2 Translating avalanche bulletins 2.1 The project The goal of the project is to implement a machine trans- lation system capable of translating Swiss avalanche warning bulletins. These are prepared in German by the IFENA (Swiss Federal Institute for the Study of Snow and Avalanches)2 a number of times a week during the winter season, the exact frequency of their appearance depending on weather conditions. They are then trans- lated into the other official Swiss languages, French ant Italian. This research project aims to provide a more rapid, coherent and systematic translation of these bul- letins. 1 “Environnement Linguistique d’Unification” 2 The project is partially supported by the IFENA. 2.2 The sublanguage of avalanche bulletins The bulletins (cf, appendix 1) are an ideal application for a machine translation system; they deal with the limited domain of avalanches in Switzerland. This domain is well known and already partially formalized in the prac- tical guide to avalanches [Salm, l982], where the general meteorological conditions and the different geographic regions of Switzerland are defined. It is thus possible to identify the special language used in the bulletins and to limit the system to the linguistic phenomena encoun- tered in this sublanguage.3 As a result lexical, semantic and syntactic ambiguities can be considerably reduced and treated more easily. In a sublanguage the number of words and their meanings are limited and the same expressions are used repeatedly. Moreover the input is controlled in a very strict way as there is only a limited number of syntactic structures and it is neither necessary nor even desirable to express the same idea in different ways [Lehrberger, 1982]. The bulletins inform the population about the snow condition and the …
Content truncated to 500 words
Labeled Text
AbstractThis presentation outlines the implementation of a machine translation system for avalanche warning bulletins in natural language, using a unification-based formalism developed at ISSCO, which will be introduced at the same occasion. Concrete examples taken from this project exemplify a modern approach to ma- chine translation: a rich representation of the semantic content of a sentence, the use of a sin- gle grammar for parsing and generating as well as generation and transfer based exclusively on the semantic representation of a sentence. Simultaneously, the limits of bidirectional trans- fer are being tested.
Publication Details
- Paper ID:
- 608
- Year:
- 1991
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan