20/11/2003 09:59:05, Stevan Harnad <[email protected]> wrote: > > >(1) The discussion was about whether there is any need for >a human-generated subject index when a full-text inverted index is >available for boolean search. With abstract/indexing services only >article titles and abstracts are available for searching, not article >full-texts.
Agreed, but I would beg to differ that we are speaking about the same thing when comparing inverted file full text with something like Medline, where indexing terms are allocated from a carefully crafted thesaurus where the medical terminology is rationally and logically structured. > >(2) Even with abstract/indexing services it would be interesting to >find out which users and how many do and do not use the subject index, >and why and why not (and how long the subject index will continue to >be a human-generated one -- if it still is at all -- in the era of >automatic tools such as latent semantic indexing and the other new >similarity and classification metrics). Again agreed, but most of the studies done so far on comparing human indexed with machine indexed texts show that the latter is not (yet?) at the point where synonyms and such like can be dealt with. In any case I think the original entry on which this thread has been predicated shows that the language in many areas is not precise enough to make unambiguous indexing possible. We are in a changing world, progress is undoubtedly being made but we are not 'there' - whatever that is, yet. Barry Mahon
