[Imports] TIGER name expansion (was GNIS tags)

Serge Wroclawski emacsen at gmail.com
Fri Sep 13 13:01:02 UTC 2013


On Fri, Sep 13, 2013 at 8:38 AM, Richard Welty <rwelty at averillpark.net> wrote:

> another error of the west coast bot that has gone unmentioned here had
> to do with single letter street names. the example i encountered was in
> Nevada, Iowa, where the bot had incorrectly expanded E St to East Street,
> N St to North Street, and so forth.

Right, that was also a big one.

Because Roy (the TIGER expansion bot) had been designed to take a very
conservative approach, it did not fix those names back to E Street. In
theory that could be done, though.

As for fixing that in the future, direction prefixes/suffixes are
hard, and while you can do some clever tricks to be more sure, you can
never be 100% sure about what the right answer is without local
knowledge.

For example, if a new version of Roy saw "N E St", it could look at
"St" and by position assume it was the road type Street. It could also
see that the name had three words (vs "E Street", which only has two),
and then the first part of that was the name of a direction "N", so it
/probably/ meant "North E Street", but E is ambigious. E, in this
case, might still mean "East", or maybe the street really has the base
name "N E". TIGER preserves this information, but without the benefit
of the TIGER information, there's no algorithm you can use to
determine this programatically. You can't even make this a MapRoulette
style challenge because a non-local wouldn't have enough information
to go on.

This is why I'm more hesitant to have Roy be redesigned to handle all
the complexities of road name expansion where a human mapper has put
in the contracted version. We can be fairly confident about "St" and
"Rd" (as long as their position is the last thing in the name), but
once you start to get into directions, things get tricky.

On the other hand, if Roy were to look at the road names and look for
incorrect expansions, it could fix those, if the road still has the
TIGER tags on it.

- Serge



More information about the Imports mailing list