Words in the natural language adapt to new meanings, lose old meanings, rearrange current meanings and change some parts of previous meanings, etc. The language as a living organism needs to be able to accept and adapt to those changes. Artificial languages such as classification systems or subject indexing systems have to adjust to those changes too. Linguist F. de Saussure defines language as a 'system of signs'. If we transfer this definition into the artificial language such as a classification system (e.g., Universal Decimal Classification (UDC)) and define it also as a 'system of signs' which has a vocabulary, a grammar and a syntax, we could draw parallels between phenomena which occur in natural and in artificial languages. The linguistic discipline which deals with changes in meanings over time is called diachronic semantics and this paper explores how its mechanisms can be used to analyze the changes that occur in classification systems over time. The mechanisms used to describe adapting and changing meanings of words include: metaphor, metonymy, specialization, generalization, analogy and splitting. This paper also aims to explain the borrowing and adjusting of the theory from the field of linguistics into the field of information sciences.