Re: MaJorToM version 1.2.0 released

Hi Lars,

no, it's not. But basically the work was as follows: Take 2 to 4 weeks
worth of the electronic program guide [1] and convert it to TMDM (in
the same naïve mapping as we did it for the AWD). Also merge the
schema stuff which comes in the epg include.zip. Connect the loosely
coupled data (by it's ids) by converting it to real associations.
Extract the "interesting" stuff, e.g. get the single actor names out
of the actors occurrence which looks something like "Tippi Hedren
(Melanie Daniels),Suzanne Pleshette (Annie Hayworth),Rod Taylor (Mitch
Brenner)" and move it e.g. to associated topics of type "Person". Then
do a NER of the people (producers, actors, movies, etc.) against a
Freebase [2] dump, maybe resolving synonym names using the connection
to the movies. By adding Freebase Ids and e.g. IMDB ids to the so-far
anonymous (because they only have a name) topics many of them will
merge. Now take everything which is reachable within 2 steps from
freebase and convert it to TMDM (this time using associations). Then
export the file(s). As all of this may be quite big if done in one
step it may be done in several steps with subsets. It's also worth
caching results from expensive requests or computations - we used
Redis for this). Depending on when you split the parts, merging them
into one big map (i.e. importing one after another) will also take
some time.

We did all of this in small steps and stored most of the intermediate
steps so we could repeat the partial steps whenever we wanted with
only "our" computing overhead. MaJorToM was improved quite a lot in
this progress and we compared with the other implementations from time
to time to see how they perform. I already wrote about the results.

Unfortunately, the code won't be publicly available for several
reasons (mostly rights of 3rd parties).

Benjamin

[1] http://epgdata.com/
[2] http://www.freebase.com/

On Sat, Apr 16, 2011 at 2:29 PM, Lars Heuer <[email protected]> wrote:
> Hi Benjamin,
>
>> In our benchmarks MaJorToM and Ontopia where always fastest, often
>> very close together. In all of our tests, tinyTiM was the slowest
>> implementation, sometimes very far off.
>
> This is a very sad news. :)
>
> Is the suite published somewhere?
>
> Best regards,
> Lars
> --
> Semagia
> <http://www.semagia.com>
>
> <https://twitter.com/larsheuer/> Twitter
> <http://www.topicmaps.de/mailinglist/> German Topic Maps mailing list
> <http://tinytim.sourceforge.net/> Open Source Topic Maps engine
> <http://mappa.semagia.com/> Mappa - Python Topic Maps engine
>
>
_______________________________________________
topicmapmail mailing list
[email protected]
http://www.infoloom.com/mailman/listinfo/topicmapmail