This has the potential to work, but we need to be careful that the
descriptions don't only partially represent their subjects. This is
especially difficult with humans, as they are often known for several
things, and occasionally (but in a statistically significant number, I
would think), known for things that don't fit cleanly into a "[nationality]
[career], born [birth year]" formula. As it exists now, the Wikidata item
on the Ft. Hood shooter, Nidal Hasan [1], gives his military branch and
rank, his location and place of birth, his gender, and a Commons category.
From that, a bot summary would likely be "American
Army major, born 1970".
There would be no indication of his source of
notability, the shooting.
What I would recommend is that we start with inanimate objects and get our
bearings on bot-generated descriptions there (celestial objects, video
games, buildings), then move onto the slightly more complicated to define
non-human living things (species of plant, species of animal, species of
creepy-crawly) and geographic locations (rivers, villages/towns/cities,
mountain ranges), and then finally onto humans.
Some things to think about: How do you create a description for a
battleship that saw service with several different navies or a river that
runs through several different countries? How do you create a description
for a country that does not exist anymore or a location that has been
destroyed? How do you create a description for a fictional person, item,
place, etc., when Wikidata does not currently have an effective way of
denoting that something is fictional? It might make sense to use Wikipedia
categories to augment the Wikidata statements.
I think that we should build a few formulas that are... difficult to screw
up. Video games come to mind, because the formula "[year of first
publication] [genre] video game" is really all you need, and other than
that some games have multiple genres, there's no way to get the description
wrong. Once the people with coding knowledge figure out what they want to
do implementation wise, I'll be happy to work with the formulas.
[1]
http://www.wikidata.org/wiki/Q1400551#sitelinks-wikipedia
On Sat, Sep 7, 2013 at 7:12 AM, Luca Martinelli <martinelliluca(a)gmail.com>wrote;wrote:
2013/9/7 Magnus Manske
<magnusmanske(a)googlemail.com>om>:
I believe that, for items that have basic
claims/statements, short
descriptions can be generated automatically, for supported languages. If
we
have "person", "Belgian",
"painter", and birth/death year, a sentence
like
"Belgian painter (1900-2000)" can be
constructed. Some awards (Nobel
prize,
Victoria cross, etc.) could be added.
+1 on the idea. Not sure about the birth/death year, though.
--
Luca "Sannita" Martinelli
http://it.wikipedia.org/wiki/Utente:Sannita
_______________________________________________
Wikidata-l mailing list
Wikidata-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-l