Towards a Bootstrapping Framework for Corpus Semantic Tagging

Availability of source information for semantic tagging (or disambignating) words in corpora is problematic. A framework to produce a semantically tagged corpus in a domain specific perspective using as source a general purpose taxonomy (i.e. WordNet) is here proposed. The tag set is derived from higher level Wordnet synsets. A methodology aiming to support semantic bootstrapping in a NLP application is defined. Results from large scale experiments are reported 1.