We present a language model consisting of a collection of costed bidirectional finite state automata associated with the head words of phrases. The model is suitable for incremental application of lexical associations in a dynamic programming search for optimal dependency tree derivations. We also

We compare the effectiveness of two related machine translation models applied to the same limited-domain task. One is a transfer model with monolingual head automata for analysis and generation; the other is a direct transduction model based on bilingual head transducers. We conclude that the head transducer model is more effective according to measures of accuracy, computational requirements, model size, and development effort.

We present a concept-based approach to the problem of lexical selection which allows us to deal precisely with lexicalization gaps or mismatches in mapping from a source language into a target language. We adopt a linguistically motivated scheme that makes use of a decompositional representation which can be classi ed in terms of a KR ontology. A main contribution of our representation scheme is the ability to handle cases where two concepts overlap, but neither subsumes the other. We have demonstrated the utility of our approach for mismatched verbs in English and German.

This paper describes a multi-level design, i.e., a non-uniform approach to interlingual machine translation (MT), in which distinct representational languages are used for different types of knowledge. We demonstrate that a linguistically-motivated "division of labor" across multiple representation levels has not complicated, but rather has readily facilitated, the identification and construction of systematic relations at the interface between each level. Our approach assumes an interlingua derived from the lexical semantics and predicate decomposition approaches of Jackendoff (1983; 1990) and Levin and Rappaport-Hovav (1995a; 1995b). We describe a model of interpretation and representation of natural language sentences which has been implemented as part of an interlingual MT system called PRINCITRAN.

Since I have already edited your manuscript it would be easier for me if you simply itemized your response on an email message rather than sending another manuscript copy. Thanks Norm Haard

This paper describes the design and implementation of an English-Turkish machine translation (MT) system developed as a part of the TU-Language project supported by a NATO Science for Stability Project grant. The system uses a structural transfer approach in translating the domain of IBM computer manuals. The general design of the translation system and a tailed description of the transfer component is presented in this paper.

are each other's translation if they are built up from parts which are each other's translation, by means of translation-equivalent rules. Principle of Isomorphism: Two sentences are considered translations of each other if they have the same semantic derivation trees (hence corresponding syntactic derivation trees). Within this framework there is a close relation between syntax and semantics: the meaning of an expression is a function of the meaning of its parts, which, in turn are defined by syntax. The translation relation corresponds to the "tuning" of grammars I For example, it might be possible to include some programming problems at the end of each chapter that would allow the student to incrementally build a mini-translation system between two languages using the principles of the Rosetta approach. An additional instructional tool would be the provision of software for use in the development of such a system. 582 across language pairs, i.e., every lexical item and rule