Evaluating specifications for controlled Greek

Evaluating Specifications for Controlled Greek Marina Vassiliou Institute for Language & Speech Processing Epidavrou & Artemidos 6, Maroussi, 15125 Athens, Greece mvas@ilsp.gr Stella Markantonatou Institute for Language & Speech Processing Epidavrou & Artemidos 6, Maroussi, 15125 Athens, Greece marks@ilsp.gr Yanis Maistros Department of Electrical & Computer Engineering, National Technical Univer- sity of Athens, Athens, Greece maistros@softlab. ece.ntua.gr Vangelis Karkaletsis Software and Knowledge Engineering Laboratory Institute of Informatics and Telecommunications, N.C.S.R. “Demokritos” vangelis@ iit.demokritos.gr Abstract In this paper we report on the set of con- trolled language specifications defined for Modern Greek and the development of the respective style checker. We will focus on the effectiveness and suitability of these specifications by assessing the perform- ance of a commercial machine translation system over controlled texts and will comment on the evaluation results. For our experiments we have used the SYSTRAN MT system (English-into- Greek language pair). We will show that an improvement in translation is feasible, when a text compliant with controlled language specifications enters a MT sys- tem. Finally, we will propose a third pa- rameter for setting CL specifications. 1 Introduction In this paper we outline the first, to the best of our knowledge, attempt to define specifications for controlled Modern Greek and develop the relevant authoring tool1, aiming at the production of texts, which make a good input to MT systems (Huijsen 1998). 1 This tool is the result of the research project “SCHEMATOPIISI: Integrated environment for the development and exploitation of Greek controlled sub- languages” (ΕΠΕΤII/98), which was funded by the Greek General Secretariat of Research & Technology. Our main concern here, though, is to test the va- lidity of the aforementioned specifications by evaluating the translations of controlled texts. To this end, SYSTRAN has been used as a testing MT platform. The linguistic and formatting specifications have been defined on the basis of the following design principles: 1. Language level: Reduction in ambiguity and redundancy together with effective terminology management 2. Formatting level: Controlled text layout, given that the text layout reflects textual structuring 3. Implementation: Use a development platform compatible to most current applications – Create a functional and user-friendly tool The evaluation procedure involved the cross- checking of the translations of two types of text: (a) "raw" texts and (b) the same texts modified to conform with the controlled language specifica- tions. The effect of the CL specifications on the translations was then measured. 2 Linguistic Specifications Basically, we have constructed a machine-oriented controlled language (cf. Huijsen 1998). A checker was implemented and the majority of the specifica- tions and instructions have been interpreted for machine use. However, some of the instructions are not exactly machine-oriented; they are recom- mendations, which allow the user a certain degree of freedom when processing his/her text. The specifications we have used here comprise both the machine-oriented instructions and the hu- man-oriented ones. 2.1 Language level By defining the language level specifications, we aimed at eliminating ambiguity and redundancy (both lexical and structural) at the morphological, lexical and …

Content truncated to 500 words