Abstract:
An approach to automatic extraction of terms from an individual scientific text is reported, which combines known methods: linguistic patterns, statistical terminological measures, methods of graph ranking. The combined methods and stages for extracting, selection and ranking of terms are described, which are implemented for processing documents in Russian. The results of experiments on extracting terms from educational texts in mathematics and programming are presented. The scores of extraction efficiency (74% of average accuracy) show that the described approach is promising.
Keywords:natural language processing, automatic term extraction, linguistic templates, graph ranking methods.