The dataset contains 22,950 term candidates extracted from 15 Slovenian PhD theses. The term candidates are of length 1 to 4, extracted via morphosyntactic patterns and the frequency threshold of 3. The PhD theses are from the areas of chemistry, computer science and political science. Each of the term candidates is annotated by four annotators as being (1) in-domain term, (2) out-of-domain term, (3) general academic term or (4) not a term. Each term candidate is also annotated with its frequency in the PhD thesis and 7 statistical measures. The resource can serve as a training set for supervised learning of term extraction and for terminology extraction tool benchmarking.