Grammatical patterns in the corpus-driven “Lexical Database of Lithuanian”

Abstract

This paper describes the grammatical patterning of two parts of speech – nouns and adjectives – included in the corpus-driven “Lexical Database of Lithuanian” as a foreign language. The lexical database is a lexicographic application of the Lithuanian Pedagogic Corpus (approx. 620.000 tokens) which was used to develop headword lists and to collect word usage information in the form of corpus patterns. In this project, we adopted a partially automated inductive procedure of Corpus Pattern Analysis for 207 verbs, 386 nouns, 87 adjectives, and 41 adverbs. The detected corpus patterns reflect different meanings of the headword. Each pattern presents information on grammatical, semantic, and lexical levels. Manually selected examples illustrate all pattern components. In this paper, 673 patterns with nouns and 99 patterns with adjectives will be analysed discussing their syntactic behaviour in detail and providing some comments on lexis-grammar interface. The majority of patterns with nouns and adjectives are minimal patterns which include only the closest syntactical partners. This result is influenced by different procedures used to describe patterns with nouns, adjectives, and adverbs and patterns with verbs. Due to rich grammatical information, there are several similar patterns with one main (usually the most frequent) type and its variants. Pattern variants show that the grammatical characteristics of a specific word usage are rather individual.