Pius ten Hacken

Has There Been a Revolution in Machine Translation

When we compare the contributions on MT in the proceedings of Coling 1988 and Coling-ACL 1998, it seems obvious that in the period between them a revolution has taken place. Often this intuition is formulated as the replacement of linguistic approaches by statistical approaches. On closer inspection, however, this position cannot be defended. An analysis of Rosetta, concentrating on the different levels of discussion and of underlying assumptions, shows that the choice of knowledge from linguistic theories or information theory and corpora is by itself not a decisive issue. More important is the question of how the problem to be solved by an MT system is defined. An analysis of the decisions underlying Verbmobil, resulting in a list corresponding point by point to the one for Rosetta, shows how far-reaching the new approach to defining the problem of MT is. As it is shown that these systems are representative of the work in MT as it was done ten years ago and today, it can reasonably be argued that a revolution in MT has taken place, though not in exactly the way it is often believed.When we compare the contributions on MT in the proceedings of Coling 1988 and Coling-ACL 1998, it seems obvious that in the period between them a revolution has taken place. Often this intuition is formulated as the replacement of linguistic approaches by statistical approaches. On closer inspection, however, this position cannot be defended. An analysis of Rosetta, concentrating on the different levels of discussion and of underlying assumptions, shows that the choice of knowledge from linguistic theories or information theory and corpora is by itself not a decisive issue. More important is the question of how the problem to be solved by an MT system is defined. An analysis of the decisions underlying Verbmobil, resulting in a list corresponding point by point to the one for Rosetta, shows how far-reaching the new approach to defining the problem of MT is. As it is shown that these systems are representative of the work in MT as it was done ten years ago and today, it can reasonably be argued that a revolution in MT has taken place, though not in exactly the way it is often believed.

Word Formation in Electronic Dictionaries

In this article I will discuss various aspects of the relationship of word formation to dictionaries. Although the focus here is on electronic dictionaries, the discussion will also touch on certain types of dictionaries for human use, especially learners dictionaries. I propose that an approach diverging significantly from the traditional treatment of word formation in dictionaries has decisive advantages for the development and use of electronic dictionaries, and that some of these advantages carry over to the treatment of word formation in learners dictionaries. In section 1 the problem is explained. Section 2 is devoted to a review of some approaches that are modeled on the traditional division of labor between dictionaries and grammars. In section 3 the approach incorporated in Word Manager is presented, which has a number of advantages compared to the approaches discussed in section 2. Section 4 outlines a way of using these advantages in learners dictionaries. Section 5 summarizes the conclusions.

The semantics of word formation and lexicalization

This is an innovative approach to word formation and lexicalisation. In the study of word formation, the focus has often been on generating the form. In this book, the semantic aspect of the formation of new words is central. It is viewed from the perspectives of word formation rules and of lexicalization. An extensive introduction gives a historical overview of the study of the semantics of word formation and lexicalization, explaining how the different theoretical frameworks used in the contributions relate to each other. Each chapter then concentrates on a specific question about a theoretical concept or a word formation process in a particular language and adopts a theoretical framework that is appropriate to the study of this question. From general theoretical concepts of productivity and lexicalization, the focus moves to terminology, compounding, and derivation. Theoretical Framewords discussed include: Jackendoffs Conceptual Structure; Langackers Cognitive Grammar; Liebers lexical semantic approach to word formation; Pustejovskys Generative Lexicon; Beards Lexeme-Morpheme-Base Morphology; and, the onomasiological approach to terminology and word formation.

A knowledge acquisition and management system for morphological dictionaries

A system for the acquisition and management of reusable morphological dictionaries is clearly a useful tool for NLP. As such, most currently popular finite-state morphology systems have a number of drawbacks. In the development of Word Manager, these problems have been taken into account. As a result, its knowledge acquisition component is well-developed, and its knowledge representation enables more flexible use than typical finite-state systems.

The Semantics of Compounding

The question of how to determine the meaning of compounds was prominent in early generative morphology, but lost importance after the late 1970s. In the past decade, it has been revived by the emergence of a number of frameworks that are better suited to studying this question than earlier ones. In this book, three frameworks for studying the semantics of compounding are presented by their initiators: Jackendoff’s Parallel Architecture, Lieber’s theory of lexical semantics, and Štekauer’s onomasiological theory. Common to these presentations is a focus on English noun-noun compounds. In the following chapters, these theories are then applied to different types of compounding (phrasal, A+N, neoclassical) and other languages (French, German, Swedish, Greek). Finally, a comparison highlights how each framework offers particular insight into the meaning of compounds.

Motivated tests for compounding

Abstract In this article I propose a system of tests for answering the question what is a compound. Since this type of question is not common in present-day linguistics, I begin by showing the importance of having a motivated answer. Then I develop a framework for the application of a definition of compounding in the identification of constructions in any language as belonging to the phenomenon of compounding and in the identification of expressions as instances of these constructions. By imposing certain conditions on a definition, the framework guides the search for an appropriate formulation. The definition found is valid for all languages. It is at the basis of tests whose general applicability is suggested on the one hand by the solution of some apparent problems, on the other hand by some examples from English and Turkish * I would like to thank Elisabeth Engberg-Pedersen, Francois Grosjean, Ellen-Petra Kester, and Cornelia Tschichold for their comments on earlier versions of this paper. Of course t...

On the Definition of Inflection

A language-independent definition is proposed for the phenomenon of inflection. The definition is based on agreement between two elements in a certain configuration. It differentiates inflection on the one hand from derivation, and on the other hand from cliticization. This is especially urgent in Natural Language Processing systems where consistency has to be maintained even when many coders are involved. The definition can also be used in any theory of morphology, and in lexicography.

Reading distinction in MT

In any system for Natural Language Processing having a dictionary, the question arises as to which entries are included in it. In this paper, I address the subquestion as to whether a lexical unit having two senses should be considered ambiguous or vague with respect to them. The inadequacy of some common strategies to answer this question in Machine Translation (MT) systems is shown. From a semantic conjecture, tests are developed that are argued to give more consistent and theoretically well-founded results.

The Nature, Use, and Origin of Explanatory Adequacy

If we want to compare the explanatory and descriptive adequacy of the MP and OT, the original definitions by Chomsky (1964) are or little direct use. However, a relativized version of both notions can be defined, which can be used to express a number of parallels between the study of individual I-languages and the language faculty. In any version of explanatory and descriptive adequacy, the two notions derive from the research programme and can only be achieved together. They can therefore not be used to characterize the difference in orientation between OT and the MP. Even if ‘OT’ is restricted to a particular theory in Chomskyan linguistics (to the exclusion of, for instance, its use in LFG), it cannot be said to be stronger in descriptive adequacy than in explanatory adequacy in the technical sense of these terms.

Revolution in Computational Linguistics: Towards a Genuinely Applied Science

Among people working in Computational Linguistics (CL) around 1990 and still active in the field now, there is a widely shared feeling that they have witnessed a revolution. This paper shows which developments are responsible for this perception and which elements are central in the actual revolution. In order to avoid terminological confusion, the concept of revolution as it is used here is clarified first. Then the development in the subfield of Machine Translation is studied in some detail. It is argued that the actual revolution consists in a shift of attention from the application of theoretical knowledge to the solution of practical problems. To the extent that this shift is representative of more general developments in the field, the conclusions can be generalized to CL as a whole.