COLLIGATIONAL PATTERNS OF TURKISH MULTI-WORD UNITS

In multi-word unit (MWU) extraction studies, most of the challenges for rich morphology languages like Turkish can be overcome by the study of how colligational filtering works in our minds, along with how statistical and collocational sorting affects the process. Based on the assumption that lexicalization of any given collocation as a MWU also requires compatibility to some lexical or morphosyntactic constraints, this study will present the morphosyntactic tendencies observed in colligational patterns of Turkish MWUs and discuss their implications on language-specific MWU filtering processes. The aim of the study is to discuss if in Turkish, associative strength is enough for a collocation to be lexicalized as a MWU or not. Another purpose of the study is to show some morphosyntactic and lexical constraints that may validate collocations to be lexical multi-word units in Turkish. The paper will also underscore the methodological perspectives of MWU identification valid for rich-morphology languages. To achieve these goals, we first extracted MWU candidates -trigrams from a 10-million-word sub-corpus of Turkish National Corpus (TNC) by using Text-NSP (Banerjee Pederson, 2011). After that, the 3-grams were annotated by using the NLP dictionary of TNC-tagger, and classified according to their colligational patterns and lexical categories of the MWU. Most frequently observed colligational patterns are argued to be morphosyntactic tendencies governing MWU lexicalization in Turkish. In this respect, the study aims to contribute to the understudied area of formulaic language in Turkish.