Advances in language engineering may be dependent on theoretical principles originating from linguistics, since both share a common object of enquiry, natural language structures. We outline an approach to term extraction that rests on theoretical claims about the structure of words. We use the structural properties of compound words to specifically elicit the sets of terms defined by type hierarchies such as hyponymy and meronymy. The theoretical claims revolve around the head-modifier principle, which determines the formation of a major class of compounds. Significantly it has been suggested that the principle operates in languages other than English. To demonstrate the extendibility of our approach beyond English, we present a case study of term extraction in Chinese, a language whose written form is the vehicle of communication for over 1.3 billion language users, and therefore has great significance for the development of language engineering technologies.

Suppletion is where the word-forms of the same lexeme have phonologically distinct stems. A study of thirty languages shows it to be surprisingly widespread, suggesting resistance to the pressure of paradigmatic levelling. While a major factor in its preservation appears to be the high frequency of the items that display it, two other factors are in operation, the type of inflectional category involved and the nature of the distribution of stems.

Russian's second BLUE term goluboj 'light blue' constitutes a well-known exception to the Berlin and Kay basic color-term typology. If other Slavonic languages do not have a second BLUE term, then the special position of Russian requires explanation; if there is evidence pointing to two basic terms for BLUE, we have a second set of data for investigating the evolution of this unusual color system. The languages genetically closest to Russian, Ukrainian and Belarusian, are examined. Findings of the list task, a simple and elegant test for psychological salience of color terms, provide strong evidence that Ukrainian and Belarusian have also evolved a second BLUE. What is interesting is that the term is not the general East Slavonic term that Russian uses, but a Polish borrowing: blakytnyj 'light blue' (Ukrainian), blakitny 'light blue' (Belarusian). This opens up the possibility that a category, in this case the second BLUE, may be borrowed separately from the basic term that denotes it. Since the category has been borrowed separately from its term, we have evidence of a different kind that the Russian color system includes a second BLUE: the category is salient enough to be borrowed separately from the term that denotes it.

We present a network morphology analysis of Russian noun stress. Nouns have a default fixed stem stress, but some nouns have nondefault stress that may deviate in a way that is determined by the form's position within the paradigm; different declensions prefer particular patterns as their nondefault choices. Membership of a particular declension, it is argued, constrains the range of possible stress patterns. Stress is represented as a hierarchy with limited deviation in terms of number and, less often, case. Indices in the declension hierarchy are addressed to nodes in the stress hierarchy. These indices correspond to rank orderings that declensions have for stress patterns. Lexical items inherit a default value for index rank but may override this. It is not possible for any override value to be given at the lexical entry, as this has to be evaluated in the declension hierarchy. The use of cyclicity in metrical approaches is considered, and it is concluded that lexical marking is still required. In addition, it is predicted that accusative forms that are syncretic with the nominative or genitive on the basis ofanimacy must have the same stress as the form with which they are syncretic. L IntroductionIn this paper we outline an approach to stress assignment in Russian that captures the relationship between declensional class and stress patterns, ignored in most current approaches to stress. This approach is expressed in terms of the framework of network morphology (Corbett and Fräser 1993) and represented in the lexical knowledge representation language DATR, developed by Evans and Gazdar (1989a, 1989b). Corbett and Fräser (1993) have already demonstrated how a network morphology treatment of Russian inflectional morphology, by virtue of its use of default inheritance, can deal with apparent morphosyntactic

No abstract