[ML4OSM] NLMaps: Query OSM in English (Also: Looking for participants in a paid annotation project)
Simon Will
gorgor at gorgor.de
Sat Mar 27 18:49:52 UTC 2021
Hello everyone,
In the course of my Master’s thesis at Heidelberg University, I am
building an English Language Interface for OpenStreetMap, which you can
see in action currently at https://nlmaps.gorgor.de/
The interface allows asking questions like:
* Is there any café in walking distance of Heiliggeistkirche in Heidelberg?
* How many playgrounds does Oxford have?
* What is the closest drinking water fountain from Musée du Louvre in Paris?
* Tell me the distance between Kremlin in Moscow and the Houses of Parliament in London!
It translates the queries into an intermediate representation using a
character-based RNN and retrieves the answer via queries to Overpass and
Nominatim. Using a character-based RNN for this might seem a bit
strange. I went with that for a few reasons, one of which being that I
don’t need to have a separate named entity recognition model for
detecting place names because they can just be constructed from the
characters in the source.
I’d be interested to hear what you think about it! So I hope it’s
alright that I am posting this here.
Also, I need to get more queries to train the system. If you are
interested to help with that, take a look at the attachted job posting.
You will get 150 € for 400 queries, which should be doable within 10
hours. If you’re not from a country with Euro, I’m sure we will find a
solution.
Another note: Especially in light of the recent discussion about more
inclusiveness in the OSM community, I know it’s not optimal that the
interface is only in English. If there is any sustained interest in it,
I’m happy working together with people to support other languages, as
well.
Kind regards,
Simon
-------------- next part --------------
A non-text attachment was scrubbed...
Name: job_posting.pdf
Type: application/pdf
Size: 17239 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/machine-learning/attachments/20210327/404db3d6/attachment.pdf>
More information about the machine-learning
mailing list