Sense distinction using computational methods in Turkish dictionaries

NLP(Natural Language Processing) refers to general name of the study elds related with processing languages by using computer-based systems. In NLP studies, dictionaries are required as lexical and semantic resources. Because in some cases, there are necessities to match the words with their correct senses for all possible words. There are some electronic dictionaries for Turkish such as "Contemporary Turkish Dictionary(CTD)" and "Kubbealtı Turkish Dictionary". However, both of these two dictionaries cover similar and redundant senses for several words. There are 86.382 words exist in CTD that written by Turkish Linguistic Society( TDK). There can be more than ten senses for a single word in some cases. By that reason, it can be hard to determine which meanings are explanatory and/or required and which of them are multiplexed needlessly. This problem of nding distinguishing senses of the words is called as "Sense Distinction Problem". The aim of this study is to simplify the sense distinction decisions by using some computational methods. In this study, we focused on to analyse the similarities of word senses by using some computational methods such as; Edit Distance, Cosine Similarity and Jaccard Index Similarity on two well-known Turkish Dictionaries Contemporary Turkish Dictionary (CTD) and Kubbealtı Dictionary (KD).