Morphology with Two-Level Rules and Negative Rule Features

Morphology with Two-Level Rules and Negative Rule Features John Bear Artificial Intelligence Center and Center for the Study of Language and Information SRI International Abstract Two-level phonology, as currently practiced, has two severe lim- itations. One is that phonological generalizations are generally expressed in terms of transition tables of finite-state automata, and these tables are cumbersome to develop and refine. The other is that lexical idiosyncrasy is encoded by introducing arbitrary diacritics into the spelling of a morpheme. This paper explains how phonological rules may be employed instead of transition tables and describes a more elegant way of expressing phono- logical irregularity than with arbitrary diacritics, making use of the fact that generalizations are expressed with rules instead of automata. 1 Introduction The theme of this paper is how to deal with the phonological or orthographic half of the problem of computational morphol- ogy, i.e., how to handle the various problems associated with the spellings of morphemes. The examples in this paper have been drawn from English orthography but it is easy to find examples from other languages where these techniques would be applicable as well. In an earlier paper [2], I presented a formalism for two-level phonological (or orthographic) rules very similar to Kosken- neimi's [8] and described how rules in that formalism could be interpreted in a computational system. There were problems with both my formalism and Koskenniemi's that could have been solved with the device of negative rule features. In this paper I discuss these problems and their solutions. 2 Historical Note The formalism described here was developed with the goal of al- lowing the linguist to write rules with similar or even identical contexts and still have a way of processing them. This stands in contrast to Koskenniemi's formMism, which, in its initial formu- lation, seemed to rule out pairs of such rules. For instance, in Koskenniemi's formalism, as originally stated, the two rules below, a:b¢==a__~ a : c ¢:==== o~__~, Would clash. Together they assert that a lexical character/a/, preceded by a sequence of character pairs c~ and followed by a sequence of character pairs fl, must correspond to both/b/and /c/on the surface. The orthographic rules described here are used in a morpholog- ical analysis system that is based on the work of Koskenniemi, Karttunen, and Wittenburg [8,5]. Its morphosyntactic compo- nent uses, instead of continuation classes, an extension of PATR 28 type rules including a device described by Karttunen [4] for han- dling disjunction. One version of this system also uses a definite- clause grammar in addition to the PATP~-type unification, and disjunction. It has been implemented in Prolog and runs on a Sun. 3 Summary of Alternative Rule For- malism The basic idea behind the notion of two-level rule (due to Kosken- niemi [8]) is that there are two levels of linguistic information to which a rule may refer. One has to do with how a morpheme is spelled in the lexicon. That is called the lexical level. The other has to do …

Content truncated to 500 words