Re: E-R model for thesauri

From: Anil Kumar Tappetla <aktappetla_at_hotmail.com>
Date: 3 Sep 2001 16:03:01 -0700
Message-ID: <4c4fce92.0109031503.2f33326c_at_posting.google.com>


Chris <ccleveland_nospam_at_dieselpoint.com> wrote in message news:<3B93C647.96D0DAAD_at_dieselpoint.com>...
> > Only if this was a non .edu question...
> >
> > > Can anybody point me in the direction of a url containing an
> > > entity-relationship model for representing a thesaurus?
> > >
>
> I'm a non-.edu type, and I'd be curious to hear a good answer. An
> obvious answer is a single flat table like this:
>
> word (not case-sensitive, not stemmed)
> synonyms (comma-separated list)
>
> Just look up the word, get the synonyms, and add them to the
> query.
>
> Surely, though, there is a more sophisticated answer...

an E-R model for thesaurus ??! you must be able to identify what an 'entity' is ? and then what that entity is 'related' to. if you treat each word as an 'entity' and 'synonym' & 'antonym' as possible 'relations' then you can have something like this ->

        "word1"  ----synonym----> "word2" ---synonym-->"word1"
           |
           |
           |            ...and so on...
         antonym
           |
           |
         "word3" <----synonym---- "word4"
           
             

apparently you cannot represent all words on your ER diagram like this !! so you
can 'cluster' words to form entities which you can possibly call 'categories' in a crude sense. then you can use the same relations 'synonym to' & 'antonym to' for generating you ER diagram though this will be equally huge. so you can probably then introduce a hierarchy of clusters or categories, that is repeat the categorization procedure one more time on the clusters of words you produced in the previous run and you can repeat this until you have reasonably small number of entities.for eg. -

             (s)                    (s) 
{"emotions"} -> {"love","sympathy",  -> {"love","care"}->so on..
                 "hate","jealousy"} (s)     |(a)
                                     -> {"hate","jealousy"}->so on...

(a) & (s) represent relations, "antonym to" and "synonym to" respectively. note that your root category is also a word.

anil. Received on Tue Sep 04 2001 - 01:03:01 CEST

Original text of this message