Main article name not picked up
Posted on: Tue, 11/18/2008 - 12:09
Pulling out the names from this URL:
http://dotnetslackers.com/articles/aspnet/Building-applications-for-Wind...
doesn't pick out the name "Tanzim Saqib" even though it appears numerous times... Got plenty of other examples if you want them.
:)
Trackback URL for this post:
http://www.opencalais.com/trackback/9737
Sorry, James, currently we don't support documents in German.
However, if a German name would appear in an english text (or any other language we currently support), it's pretty likely we would have it extracted. Try it :)
Thanks,
Hi Gil,
Just tried it from this url: http://geekswithblogs.net/dtomicic/archive/2007/12/08/117507.aspx
(same name, but in an english doc) ... doesn't seem to pick it up either.
Thanks
James
In this specific case, there are not enough contextual evidences to tell this is a Person's name. In fact, the full name you're interested in appears only in the title and later on under the "blogs" section, and not at all in the text itself.
In addition, when Calais is presented with an HTML, it first cleans it from any non-text segments (in order to avoid commercials etc...). In this case both appearences of the full name are "lost". Note that the title of the page (the one appearing on the browser frame) is different from the title where the name appears.
I hope it's clearer now, thanks for your questions,
hey Gil,
Thanks for the further info! We're actually not submitting HTML though. I've just cut it back to a basic repro:
"On behalf of INETA Europe James Crowley would like to invite you to participate in a 1st European Silverlight Challenge!"
on viewer.opencalias.com picks up my name, but putting in Damir Tomicic instead does not. Does Calias need different levels of contextual information depending on whether the name is "known" or not, or something?
Thanks!
James
Calais indeed uses different types of contextual evidences. Very known first names like yours, are sometimes enough :)
Hi Gil - just found another one which looks like it should have enough contextual information to include it? Pasting the text from this page: http://geekswithblogs.net/silverblog/Default.aspx into calais viewer misses the name Jacek Ciereszko. I know it's not an english name, but would you expect calais to pick it up anyway?
Thanks!
It actually has nothing to do with the fact that it's not an english name, but it's just that there is not much contextual evidences in this case either. In general - OpenCalais will work best on complete text chunks, e.g. Press Releases, articles from the news etc... it is less eficient while wroking on blogs and other such sources. The latter usually have different language and style to them, that is not OpenCalais' field of expertease at the moment.
Thanks for your feedback,
Another one:
http://blogs.msdn.com/saveenr/archive/2008/11/14/consuming-extension-met...
doesn't pick up "Saveen Reddy" as a person.
James,
From a quick glance it seems that in your first example, the person name was misidentified as an Organization (Tanzim) -- probably because there is such an organization.
In the second example it seems that there isn't sufficient context to recognize Saveen as a person's name (it is mentioned only once).
I will forward these examples to our development team and we'll see how these issues can be corrected.
Michal
James -
The first example appears to be corrected. Tanzim Saqib is now recognized as a person.
Great, thanks Fran! Just another one for you,
the text from http://tomicic.de/2008/11/24/PetitionStopptWham2008Edition.aspx pasted into calais viewer doesn't pick up "DAMIR TOMICIC" as a name. I realise this is a german name, so not sure how good your support is for this, but just thought I'd let you know!
James