Improving term extraction by combining different techniques

Two different reasons suggest that combining the performance of several term extractors could lead to an improvement in overall system accuracy. On the one hand, there is no clear agreement on whether to follow statistical, linguistic or hybrid approaches for (semi-) automatic term extraction. On the other hand, combining different knowledge sources (e.g. classifiers) has proved successful in improving the performance of individual sources on several NLP tasks (some of them closely related to or involved in term extraction), such as context-sensitive spelling correction, part-of-speech tagging, word sense disambiguation, parsing, text classification and filtering, etc.

In this paper, we present a proposal for combining a number of different term extraction techniques in order to improve the accuracy of the resulting system. The approach has been applied to the domain of medicine for the Spanish language. A number of tests have been carried out with encouraging results.