[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] Keyword Indexing

Murray Altheim wrote:    (01)

> cdent wrote:
> >>There's a CMU system I used at NTTC that
> >>could analyze a text and come up with a set of keywords for it.    (02)

The more I think about this, the more I realize that extracting keywords
from text is, in essence, the same as a search engine.    (03)

There are two ways in which such a system can provide benefit
however.    (04)

1. The ability to define synonyms.
    When equivalences are created, then all of the messages pertaining
    to the early version of X (when it was called Y) can be found.    (05)

    Similarly, improve == improvement == improving == bettering ==...
    (you get the idea -- shades of cyc) can help find important,
    useful information that otherwise would not be found.    (06)

    The thing is, the definition of synonyms is something that could be
    added to a search engine over time. A special keyword that acts
    like OR, as in (x SYN y SYN z) would add the synonyms to the
    search engine's glossary.    (07)

2. The ability to specify meta words.
    Meta words are categorziations that are *not* found in the text.
    For example, [+] for an argument for, or [Query]. That makes it
    possible to build up an IBIS structure.    (08)

I note, however, that message-level granularity is still WAY to high,
for reasons that have been discussed at extreme depths a couple
of years ago, now. The "instant outlining" approach, coupled with
purple numbers and the categorizing capacities described above
seems to me to offer a degree of promise.    (09)