[Talk-us] More road name expansion thoughts
Serge Wroclawski
emacsen at gmail.com
Sun Jul 20 01:54:26 UTC 2014
Thanks to all the folks who've commented on this thread and also to
the folks who contacted me off list about this. A majority of the
feedback I received was very positive, which I'm thankful about. I
attribute much of this to the very conservative approach that was
taken last time, because there were so few problems, people are more
open to more of the same kind of work.
I want to give a comprehensive answer to a lot of your questions.
* Problems with bad TIGER data
I heard that there were a few questions raised about the last bot
expansion. It turns out that there were problems last time related to
bad TIGER data. So for example, the TIGER data may have included
something indicating that the road had a directional suffix for East,
and the road name was Foo T E, since the road had a TIGER tag
indicating the directional suffix, the name was changed to Foo T East.
The previous bot used the TIGER data to do the name expansions. If
TIGER was wrong, so was the name expansion.
* Missed contractions in the previous name expansion
I received some feedback about lesser used contractions not getting
expanded. This should probably be addressed. I'm collecting these and
will likely propose just quickly running through them, outside of this
expansion.
* Common contractions
People seemed pretty to agree that for a majority of cases, if the
contraction is either a prefix or suffix and is relatively
unambiguous, like "Rd" or "Blvd", we can just expand it. I think this
also extends to "NW" and other two letter directions.
* Concerns about ambiguous contractions and local editing
A few people brought up a concern about words which could be
contractions but aren't always, such as "E" or "S". I agree with this
concern and I think we need to put the plain ordinals into a separate
cateogry.
I agree and I also agree with people who expressed concerns that the
only way to solve these problems is through direct survey. Maybe the
solution here is to create notes?
* Continually running bot
There was discussion about a continually running bot that would go
around and fix these, as well as other problems. This is something
that OSM has has in the past with fixbot, xybot and currently has with
WALL-E in Germany. I think it might make sense for us to have
something similar here in the US to address common problems, like
expanding Rd to Road, or "road" to "Road".
I'm in favor of this idea, but I'd like to hear more feedback about
this. If we did it, I'd want the process to be as transparent as
possible.
- Serge
More information about the Talk-us
mailing list