Lexical Semantic Change, the temporal evolution of the mapping between word forms and concepts, can be studied under two complementary perspectives: semasiology studies how given words change in meaning over time, while onomasiology focuses on how some concepts change in how they are lexically realized. For the most part, existing NLP studies have taken the semasiological (i.e. word-to-concept) view. In this paper, we describe a novel computational methodology that takes an onomasiological (i.e., concept-to-word) view of semantic change by directly inducing concepts from word occurrences at the different time stamps. We apply our methodology to a French diachronic corpus. We examine the quality of obtained concepts and showcase how the results of our methodology can be used for the study of Lexical Semantic Change. We discuss its advantages and its early limitations.