[Freebase-discuss] Metadata capturing tools & ideas for enhancing Freebase's structured data
narphorium at gmail.com
Wed Jul 7 01:24:25 UTC 2010
Wow, that's so cool. Did you really just build that all now? I really
like how dragging and dropping the Wikipedia links lets you avoid the
whole reconciliation song and dance. When I think of what the web would
look like if more people adopted structured data, being able top
drag-and-drop topics instead of copying-and-pasting text is something
I'm really looking forward to.
Also looks like my code could use a little freshening up. I'm using
Calais to do the NER but it seems like they've changed a bit since I
Spencer Kelly wrote:
> so i took a shot at making this
> its basically a clone of an app shawn wrote last year -
> the idea is to let you drag links right onto the freebase properties,
> and it writes them in.
> lots of work to do !
> On Tue, Jul 6, 2010 at 1:03 PM, Thad Guidry <thadguidry at gmail.com> wrote:
>> Freebase's search is currently lacking full-text indexing, as I understand
>> it, primarily because of lack of hardware resources. (If I'm wrong, speak up
>> and let me know otherwise) It would require more hardware & storage to
>> duplicate the index of wikipedia content that is held by Freebase. How much
>> more resources, I don't know. But for structured data, it's less useful to
>> index EVERYTHING as we know, and instead what's more useful is capturing
>> facts and metadata, the basic premise of Freebase. So how could we capture
>> more metadata from external sources ?
>> A thought:
>> (A slicker manual human labor process, since work is always progressing on
>> other machine automated ones.)
>> My vision would be an Acre app or webapp that would let you choose a
>> structure (/company_name, /person, whatever) and then using the full-text
>> index results from dbpedia or wikipedia to drive your discovery process and
>> let you highlight /text (the metadata) in paragraphs that you wanted to
>> capture and assign to Types, Properties, etc, and perhaps even reconcile
>> with existing ones. Why do I have to use dbpedia or wikipedia results to
>> drive my discovery process? Because as stated that would require more
>> hardware from Freebase and Metaweb to index, and we would basically be
>> duplicating dbpedia and wikipedia's efforts on the full-text search front,
>> and it's not what Freebase is all about. It's about structured metadata and
>> LinkedData, and a great community.
>> However, I don't think I've seen a nice metadata capturing tool for Freebase
>> built yet. One that would show wikipedia or dbpedia results on one side of
>> the browser window or even a separate tab or window and Freebase metadata
>> capturing tools on the other side or a different window. I've seen some
>> REALLY nice ones however in the Enterprise. Better tools such as this would
>> allow capturing more metadata, which we all know lends itself to the power
>> that we have with Freebase. Eventually NER technology could be utilized
>> with borrowed ontology techniques (see 1) and additional LinkedData and
>> implicit schema-level relations (http://bit.ly/daGJjr) to make Type and
>> Property assumptions and let humans refine and pick the best ones.
>> Not something like
>> this: http://www.esri.com/news/arcuser/0410/graphics/myenvironment_3-lg.jpg
>> But something more like this as a really, really crude starting
>> point: http://bit.ly/c0NUGF
>> (1) Ontology Tools
>> (2) Visualization Tools
>> You are receiving this message because you are subscribed to the
>> Freebase-discuss mailing list.
>> To post a message to the list: Freebase-discuss at freebase.com
>> To unsubscribe, view archives, etc:
> You are receiving this message because you are subscribed to the Freebase-discuss mailing list.
> To post a message to the list: Freebase-discuss at freebase.com
> To unsubscribe, view archives, etc: http://lists.freebase.com/mailman/listinfo/freebase-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Freebase-discuss