# Adapting to fluid language use

The advantages of my benefiting from your experience training a system are obvious. Just as obvious, though, are ways in which you and I might differ in our notions of relevant documents. Even this comparison is only possible if we each evaluate exactly the same query, and how often will that happen?!

If we are to change our representations of documents based on user's opinions, should we value expert" opinion over those of novice?" Common sense would suggest that if we could ascertain that one user was indeed expert in a particular topical area, then their assessments of relevance should perhaps carry more weight. But based at least on the experience of lawyers searching case law [REF1064] [REF122] experts are less likely to search in their own area of expertise, probably because they already know what they will find there. The typical user searches in areas they know less about. In that respect, perhaps the novice's RelFbk is, in fact, a better characterization of what we should attempt to satisfy?!

Once the index representation is allowed to change in response to users' we are faced with the question of just how fast this change should occur. Some of these questions were mentioned in terms of conceptual drift by users and topic tracking in news sources (cf. Section §7.3.2 ). Adopting a longer term, archival perspective of a librarian perhaps, how quickly should we wish that our adaptive system track current/trendy terms? (cf. {\bf Wired} magazine's Tired/Wired Memes feature). When old terms of art have been rendered obsolete, how can we nevertheless maintain an archival record of this previous terminology? was not used to describe this area until some time later. It is interesting to wonder how an IR system might nevertheless be able to retrieve these old documents in response to a query for ETHOLOGY.}

FOA © R. K. Belew - 00-09-21