Abstract
This study deals with the recently proposed concept of so-called Context Specificity of Lemma (CSL). CSL is based on the word embedding technique called Word2vec which enables measuring lexical context similarity between lemmas.
Specifically, a recently proposed method Closest Context Specificity (CCS) is applied to a diachronic analysis of Czech texts. This method expresses how unique is a context within which a given lemma appears.
The aim of the paper is to study what kind of semantic features can CCS detect and how useful could CCS be in a diachronic semantic analysis. The second goal is to observe the relation of CCS to frequencies in the corpora.