[Developers] looking for people

Tom Morris tfmorris at gmail.com
Mon Mar 30 15:44:48 UTC 2009


John - I usually submit bug reports for any problems that I can
characterize reasonably well.  I think some improvements have been
made to a separate version of search which runs on the Communist
(oops, make that Socialist) part of the site, but they weren't, to the
best of my knowledge, available through an API anywhere at the time I
was doing my testing.

Some specific things that I think would help searching for people include:
  - better handling of name variations without requiring aliases to be
entered in the database for all of them
    (honorifics, middle names/initials, generational indicators)
  - indexing birth and death dates - they're a key identifying factor
  - better extraction and/or indexing of names from Wikipedia articles
- they almost always have the full name of the person in the lead
sentence and sometimes have aliases as well.  I can't judge whether or
not the "anchor text only" policy is the right one in general, but in
the case of WP articles about people it leaves a lot of valuable
information on the table

I'm not sure what data sets you guys use for testing, but the
Congressional Bios might make an interesting addition if you don't
already have them.  It's about 12,000 names, all of which should exist
in Freebase.  I'll see about getting my name resolver/data loader put
on Google Code somewhere so that folks can use it as the basis for a
study if they're interested.

Tom

On Mon, Mar 30, 2009 at 4:38 AM, Faye Harris <faye at metaweb.com> wrote:
> Hi,
>
> When it comes to including middle initials in searching for people in Freebase, the problem oftentimes is simply the lack of data on middle initials.
>
> In order for /api/service/search to find a person by query "Foo A. Bar", that string has to exist in Freebase as either a 1) topic name, 2) alias, 3) prefix match for name (where omissions are allowed), or 4) Wikipedia anchor text. If this person only exists in Freebase as "Foo Bar", without any alias or a Wikipedia anchor text pointing to "Foo A. Bar", then using the latter will not find the former. You can, however, find a person named "Foo A. Bar" in Freebase with a partial string such as "Foo Bar" (or "A. Bar", "Foo A", etc.).
>
> I understand that it may be frustrating to have to disambiguate search results as an extra step when a "first-name last-name" search on people can return more than a single match. I'm hoping that as the Freebase community adds more data, and as the APIs improve, problems like this will occur less and less frequently.
>
> -- Faye
>
>
> ----- Original Message -----
> From: "Tom Morris" <tfmorris at gmail.com>
> To: "For discussions about MQL, Freebase API and apps built on Freebase" <developers at freebase.com>
> Sent: Sunday, March 29, 2009 8:37:04 PM GMT -08:00 US/Canada Pacific
> Subject: Re: [Developers] looking for people
>
> "Also known as" in the web client doesn't come from the 'name' field,
> but from the 'alias' field.
>
> Finding people by name in Freebase takes a lot of work.  I recently
> loaded up the Thomas IDs for all the U.S. congress folks and almost
> all of the work was in finding the right people to work with (and the
> search API isn't much help either).
>
> Tom
>
> On Sun, Mar 29, 2009 at 11:07 PM, Milan Andric <mandric at gmail.com> wrote:
>> Hello,
>>
>> I am trying to get details on people and it has been working well
>> until I ran up against a case that only returns an empty list.
>> The query looks like :
>>
>> query  = [{
>>        "name~=":'Bobby L. Rush',
>>        "name" : [],
>>        "type"  : "/people/person",
>>        "gender" : None,
>>        #"*":[{}],
>>        'place_of_birth':None,
>>        'date_of_birth':None,
>>        'profession':[],
>>        'key':[],
>>        'guid':None }]
>>
>> Problem is I get nothing ... yet when I do a search on freebase.org I
>> get him front and center.
>>
>> http://www.freebase.com/search?limit=30&start=0&query=Bobby+L.+Rush
>>
>> Ideas on what I am doing wrong?  I tried using name as well as name~=
>> ... same result, None.  I'm using the freebase python lib.
> _______________________________________________
> Developers mailing list
> Developers at freebase.com
> http://lists.freebase.com/mailman/listinfo/developers
> _______________________________________________
> Developers mailing list
> Developers at freebase.com
> http://lists.freebase.com/mailman/listinfo/developers
>


More information about the Developers mailing list