[Talk-se] OpenStreetMap data to georeference Swedish texts

[Thomas Fischer]

Hello,

my name is Thomas Fischer and I am a researcher and teacher at the 
University in Skövde (Högskolan i Skövde).

In the past months, I have worked on a project that makes use of 
OpenStreetMap data to geolocate/georeference text phases as found in 
news paper articles or police reports. For example, if a news report 
is about a traffic accident on E20 near Vårgårda, the software's 
task is to identify the place were E20 passes closest to the city of 
Vårgårda. Similarly, if the report talks about Skolgatan, the 
software shall identify the right instance of a road of this name if 
the report names the municipality in the same input text. The 
software is designed specifically for Sweden (hierarchy of road 
names and administrative structures) and the Swedish language (stop 
words, for blacklisting certain terms, converting plural forms into 
singular, ...).

A demonstrator of the resulting software is available at
 http://fish.research.nsa.his.se:9876/

The demonstrator's source code is available at GitHub:
 https://github.com/thomasfischer-his/pbflookup

In my experience, the software is working already quite well. There 
are two main limitations:
- Missing, inconsistent, or erroneous information in the underlying
  map data from OpenStreetMap (unfortunately)
- Lack of a deeper textual analysis: Words are extracted as is from
  the input text, so the word 'vara' (to be) and the city of Vara
  are indistinguishable.

The software is written in C++, but care has been taken for code 
quality by using static code checkers and tools like Coverity Scan 
or Valgrind.
The project to develop this softwas was co-financed by 
Internetstiftelsen

I am contacting this mailing list for two reasons. First, I would 
like to get some feedback such as error reports or ideas for 
extensions.
Second, I would like to pursue this project further, so I am looking 
for cooperation partners that can (financially) support the project. 
Examples for cooperation partners can be app developers that want to 
make use of the JSON/XML interface, news agencies or newspapers that 
want to add the service of matching their articles to a map 
location, or governmental, regional, or municipal agencies that want 
to geolocate their own or citizens' reports (think of 112 calls or 
citizens reporting broken street lights).

So, if you have an existing contact at an organization that may be 
interested, please forward this mail. If you personally interested 
in a cooperation, you can contact me directly, of course.

Greetings from Skövde, Sweden,
Thomas Fischer
https://www.his.se/fish
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 163 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.openstreetmap.org/pipermail/talk-se/attachments/20160915/8867c050/attachment.sig>