Interchangeability as the feature of synonymy (experimental and corpus research)

Abstract

In the paper interchangeability is regarded as a feature (criterion) of synonymy. The research is based on the psycholinguistic experiment and Russian National Corpus data. 144 subjects took part in the experiment: they were given 22 sentences in which subjects could choose one of the given synonyms. The hypothesis of the research is that the interchangeability depends on the types of semantic relation within synonymic groups; synonymic relations are directly organized. The important task of the research is to determine the role of collocation in the choice of synonyms. MI (Mutual Information) is а special tool to measure the collocation strength, it measures a relationship between two random linguistic items that are sampled simultaneously. With the help of the Russian National Corpus data, MI is calculated for each utterance. There are few types of synonymic groups with different semantic relations: centric, middle asymmetric, and non-centric. These types are distinguished in my previous paper. The centric synonymic groups have the headword (typically the headword is the most frequent word); the middle asymmetric synonymic groups have two headwords, that are the core synonyms; the non-centric synonymic groups have no headword. Directivity of synonymic relations depends on the type of a synonymic group. It is possible to introduce three following models: for the centric synonymic groups: subordinated synonym → headword; for the middle asymmetric groups: subordinated synonym → core synonyms; for the non-centric synonymic group: synonym → another synonym. As a result, it was shown that different contextual features are significant (for example, the meaning of neighboring words, general utterance sense) for the choice of synonyms.