Automated Identification of Domain Preferences of Collocations

This paper addresses (semi-)automatic collocations dictionary compilation in connection with the automated identification of domain preferences of collocations. The research was motivated by the process of the semi-automatic compilation of the Estonian Collocations Dictionary (ECD), where lexicographers processed a large number of terminological collocations extracted from Sketch Engine into the Dictionary Writing System EELex. In this paper, we apply the terminology extraction module within the Corpus Query System Sketch Engine and present the results of the experiments on building military domain corpora in Russian and Estonian and extracting multiword terms. Both languages have very rich morphology and quite a large number of multiword terms, but Russian texts are well represented on the Web while Estonian ones are not. We analyze how the comparison of frequency of a collocation in a reference corpus with its frequency in a domain corpus can be used for facilitating word sketch data analysis in terms of identification of domain preference of collocations.