IN THE QCFQBER 2O, 2000 ISSUE, I DIPPED MYTOE INTOTHE LARGE LAKE OF TEXT STRING searching. I described two approaches for handling a list of keywords describing an archive of documents. But a keyword-based approach to accessing a large number of documents makes some strong assumptions. A really good set of keywords may require human reviewers, a fact that certainly restricts the scale of the document archive. There aren't enough human reviewers to index millions of documents. Some interesting progress has been made in automatically generating keywords for documents, much like you can automatically generate book indexes. But even if you have a pretty good set of keywords describing documents, your end users are still left with some tricky query issues.
展开▼