Lingvisticæ Investigationes - Volume 21, Issue 2, 1997

In this paper, we examine the adjectives that seem to require two arguments in symmetric relation. To syntactically define these elements, we observed whether an adjective can appear in the following transformational relation or not: N0 N1-wa Adj = N1 N0-wa Adj. Those that allow this relation are classified in the class AWS and named Symmetric Adjectives. They constitute a semantically and syntactically homogeneous class. Two types of problematical structutures were considered: one is the construction containing a complement in -ei, replaceable by a symmetric complement in -wa; the other is the construction with complement in -wa which is not a symmetric one. Besides, we came across one more complicated structure concerning symmetric construction, that is, N0 N1-wa N2-i Adj. However, it cannot be considered as the basic structure for symmetric adjectives, since N2-i is either a restructured element or a symmetric noun: in the first case, this structure is a derived one; in the second case, it is this noun that requires a symmetric complement Nj-wa, and not the adjective. Therefore, this structure cannot characterize symmetric adjectives. In this way, we obtained 120 adjectives for the class AWS among 5300 Korean simple adjectives. The result of our study is represented in the form of a matrix in the Appendix. More detailed studies on symmetric verbs and nouns are also required as well as more advanced analyses about semantic and logic constraints in the distribution of arguments.

This article presents a first attempt at systematically classifying neutral verbs in English. Neutrality, also known as the ergative construction or the causative alternation, is present when the following equation holds:(1) N0 V N1 = N1 VIn the following sentences, for example:(2) a. Max chimed the bells.b. The bells chimed.(3) a. The enemy sank the battleship.b. The battleship sank.the direct object of the (a) sentence is the subject of the (b) sentence, and the meanings are similar. Following Boons, Guillet, Leclère (1976), who presented a classification of over 400 neutral verbs of French, we systematically examined the English lexicon and compiled a list of over 500 neutral verbs of English, indicating an appropriate N1 along with the morphological properties of human and non-human for each N0 and N1. A preliminary glance at these verbs suggests that the morphological and transformational properties associated with each verb are highly idiosyncratic. Furthermore, this study shows that although certain semantic classes tend to favor neutrality, they do not automatically assure it, and presents further evidence for building a formal lexicon, or lexicon-grammar, where pertinent lexical and syntactic information is listed for each individual verb.

Compound nouns are independent lexical units and, therefore, they can not be directly translated on the basis of the translation equivalents of their constitutive units. In this article we propose a typology of translation divergences for compound nouns on the basis of a corpus of 25,000 translation equivalents French-Spanish. This typology take into account, on one hand, the 'transparence' degree in translation and, on the other hand, the differences between a French compound noun and its Spanish equivalent concerning the morphological form of this compound, the gender and number of its constituent units and the determination. It was observed that more than 25% of equivalent units in translation have some kind of divergence.

The historical corpus of Hungarian was collected to serve as the source of the Historical Dictionary of Hungarian which is being compiled in the Department of Lexicography and Lexicology of the Research Institute for Linguistics. In order to be able to retrieve the headword lemmas from the corpus a morphological analyser software was designed and applied. The paper describes the process of the analysis and the current results.

Although many frequency dictionaries of French words have been published over the past decades, they rarely include fixed expressions and idioms. I therefore view it as an important task to try to compile a frequency dictionary of French locutions verbales (fixed expressions of the predicate type). A pilot study of the frequency of about twenty expressions in the newspaper Le Monde 1996 and 1997 reveals that particular problems arise in creating such a dictionary. This paper discusses some of these problems, including variation (instability) and ambiguity, and concludes with a model entry for the dictionary.

This study is the first comprehensive, data-based examination of subject raising in Portuguese, and is based on 4500+ tokens in more than 26,500,000 words of text from both the written and spoken registers of Brazilian and European Portuguese. We have suggested that there are important differences in raising between the spoken and written registers, which are related to presumably universal production strategies for the two registers. Evidence suggests that morphological factors such as subject-verb agreement play an important role in determining whether raising occurs with first, second, and third person subjects. In terms of differences between the European and Brazilian dialects, we find that split agreement (eles parece saberem) and obligatory coreference {me parece ver um fantasma) are both more common in European Portuguese. Finally, these last two facts, along with a number of related phenomena, suggest that there are important differences in the underlying clause structure of European and Brazilian Portuguese, which can further be extended to include other languages such as Spanish.

In this paper we present some of the main results obtained within the European project COPERNICUS 621 "GRAMLEX" aiming at development of basic resources, tools and methods useful in language engineering. Results presented here were obtained by the Polish team of the GRAMLEX project using Polish data but are easily portable to all Indo-European languages. They are of special interest for highly inflectional languages (e.g. for all Slavonic languages). The high portability of results follows from the proposed, very general format of morphological description (GRAMCODE) and from the general character and great simplicity of algorithms referring to this format. In the paper we briefly present the GRAMODE format, as well as the fundamental dictionary based tool for morphological lemmatization and tagging of texts (LEXAN).