Automatic terminology extraction from texts is a very useful task for users like researchers in literature. Tools proposed to perform this extraction require texts tagged with grammatical categories. When dealing with texts written in Old French, this tagging is not an easy task. In this paper, we thus address the problem of tagging such texts. We consider the combination of two existing tools for tagging words: TreeTagger and Brill. We propose the so-called TERVOTIQ system (TERminologie par VOte selon l'éTIQuetage), which aims at enhancing the quality of the terminology extracted by improving the tagging. This tool is based on the use of a criterion of reliability in order to consider those terms that were extracted from the texts treated by the two taggers. Our approach allows the ordering of terms by the value of the criterion of reliability (the higher value of the criterion represents the more relevant term), and by the number of occurrences. Experiments have shown that the terms that we extract using our tool are more relevant than the ones extracted without using the combination of the other two taggers.