FAQFinder with sense tagging FAQFinder without sense tagging

Rejection Recall FAQFinder with sense tagging FAQFinder without sense tagging Figure 4: Recall vs. Rejection for FAQFinder with and without WordNet Sense Tagging search. In FAQFinder, sense tagging and calculation of semantic similarity are much more computationally intensive than term vector processing. However, since FAQFinder matches single questions rather than entire documents, the computational complexity has not been an issue. Extending our approach to matching of HTML documents would probably only be feasible if we extracted key portions of documents, such as titles or portions of the body marked with certain HTML tags.