EN
The paper presents a new corpus-driven method for automatic term recognition (ATR) based on data-mining techniques. In our research, ATR is seen as a valuable resource for the theory of terminology. The main goal of our approach is not the recognition of terms in given academic disciplines but rather a contribution to the definition of the notion “term”. The basis for such definition improvement is provided by feature-ranking, which is a built-in function of the data-mining tools and is able to determine the impact of individual statistical and linguistic features of word-forms on the accuracy of the ATR method. One of the basic characteristics of a term, its terminological strength (or termhood) depends on the combination of the most significant features in a given word-form.