[topicmapmail] automated generation of topic map

[jean delahousse]

Hello,

1) the set of topics is not pre defined but the tools only explore a certain
set of Reuters news (example : all reuter news about a specific industry or
market) and recognize company names by the context of the sentence. To get
the tools more precise it's possible to give it a first list of company
names to look.

2) the tool only searches for a limited set of association, as the tool is a
semantic tools that will try to find those association analysing english or
french sentences.

That type of solution means a heavy and fastidious work to train the tools
to recognize some specific associations in a specific language. It's ok as
this work can be reused and that in a specific domain like economical
intelligence you will try to find a finite number of association.

Anyway: it's a strong solution with no miracle

Sincerely

Jean Delahousse
Mondeca
CEO
3, Cité Nollez. 75018 Paris
Phone : 33 1 44 92 35 03
Mobile : 33 6 16 90 73 95
Fax : +33 1 44 92 02 59

www.mondeca.com

> -----Message d'origine-----
> De : topicmapmail-admin@infoloom.com
> [mailto:topicmapmail-admin@infoloom.com]De la part de Dan Wu
> Envoyé : vendredi 8 février 2002 20:41
> À : jean delahousse
> Cc : topicmapmail@infoloom.com
> Objet : RE: [topicmapmail] automated generation of topic map
>
>
>
> Hello Jean,
> Thank you for your information.
> I understand from domain expert, you could have a predefined set of
> topics.
>
> I further would like to know whehter you have a predefined set of
> association types or you can mine them from web site.
>
>
> Thanks,
> Dan
>
> > Dear Dan,
> > We have experienced for a client a solution with automatic
> feeding of the
> > topic map.
> >
> > The taxonomy for classification of topics have been made by the experts.
> >
> > The day to day feeding of the topic maps (topics, associations and
> > occurences) has been automated. A text mining tools is used to
> extract the
> > topics and associations from Reuters news. In this application the text
> > mining tools was trained to capture company names, and various
> association
> > between topics like "acquisition" "selling" "partnership".
> >
> > The automatisation using text mining will be completed soon
> with the use of
> > the new's metadata (NewsML format).
> >
> > You can see a prototype of the solution at the following adress :
> > http://1.mondeca-publishing.com/i/editortemis/title49.html
> >
> > sincerely
> >
> > Jean Delahousse
> > Mondeca
> > CEO
> > 3, Cité Nollez. 75018 Paris
> > Phone : 33 1 44 92 35 03
> > Mobile : 33 6 16 90 73 95
> > Fax : +33 1 44 92 02 59
> >
> > www.mondeca.com
> >
> > > -----Message d'origine-----
> > > De : topicmapmail-admin@infoloom.com
> > > [mailto:topicmapmail-admin@infoloom.com]De la part de Dan Wu
> > > Envoyé : vendredi 8 février 2002 17:11
> > > À : topicmapmail@infoloom.com
> > > Objet : [topicmapmail] automated generation of topic map
> > >
> > >
> > > Hello,
> > > I was thinking that constructing a topic map for
> > > a domain is a very time consuming and labour intensive
> > > job cauae one, who I assume should be the domain expert,
> > >  has to find out
> > > all the topics, occurrences and
> > > associations. Is there a way 	already devloped or being discussed
> > > that we can automate this process.
> > >
> > > Thanks,
> > > Dan
> > >
> > > _______________________________________________
> > > topicmapmail mailing list
> > > topicmapmail@infoloom.com
> > > http://www.infoloom.com/mailman/listinfo/topicmapmail
> > >
> >
> > _______________________________________________
> > topicmapmail mailing list
> > topicmapmail@infoloom.com
> > http://www.infoloom.com/mailman/listinfo/topicmapmail
> >
>
> _______________________________________________
> topicmapmail mailing list
> topicmapmail@infoloom.com
> http://www.infoloom.com/mailman/listinfo/topicmapmail
>