tailieunhanh - Báo cáo khoa học: "MORPHOLOGY in the EUROTRA BASE LEVEL CONCEPT "

Finnish by establishing correspondences between a surface alphabet and a lexical alphabet (the two levels) and using a lexicon to determine which combinations of characters and morphemes are legal. Moreover, this is done by means of declarative rules, thereby avoiding the procedural problems of generative phonology, and the algorithm used is language independent. | MORPHOLOGY in the EUROTRA BASE LEVEL CONCEPT by Peter Lau and Sergei Perschke Commission of the EC Bât. JMO L - 2920 Luxembourg ABSTRACT In recent years the nature and the role of a morphological component in NLP systems has attracted a lot of attention. The two-level model of Koskenniemi which relates graphemic to morphological structure has been succesfully implemented in the form of finite state automata. In EUROTRA a solution which combines morphological and surface syntactic processing In one CFG implemented in a unification grammar framework has been tried out. This article contrasts these two approaches considering especially the feasibility of building morphological modules for a big multilingual MT system in a decentralised R D project. o. INTRODUCTION The development of sophisticated NLP applications has created a need for specific processing in order to be able to cope with large vocabularies without creating monstruous dictionaries. Earlier approaches often avoided morphology more or less by listing full wordforms In the dictionary or by simply segmenting some Inflectional endings with a few general rules. Much recent work Is based on the Two-level Model Koskennleml 1983 and relates directly or Indirectly to the original Implementation of this model In the form of finite state transducers FST . The original notation and implementation have been further developed and refined cf. . Black 1986 and Bear 1986 In order to Improve compilation and runtime debugging and rule-writing facilities. Still some problems persist and others have not been touched yet. This paper presents an alternative but not contradictory solution which has to some extent been tried out in the EUROTRA Machine Translation Project and argues that the two-level approach may not be entirely viable In a decentralised R D project which aims at the creation of a big multilingual MT system. I. THE TWO-LEVEL MODEL The original presentation of the model Koskennieml 1983 shows that it Is .