Tuesday, December 4, 2007

GeoCoding TTC stations

Mapster is back alive and kicking!
The current item on my agenda is adding TTC Buses which involves a lot of changes to Mapster internals. Also, adding the TTC Buses requires me to parse all the data from the ttc website.
This is by far the worst data I have seen or imagined. I have a strong suspicion that all the data was entered by a blind person in 1965.

I ll write later on the assumptions that allowed me to parse the schedules more or less accurately. But in this post, I ll talk about geocoding stations. The geocoder remained unchanged ever since Ivan worked on it more than 8 months ago. It works like a champ for stations that are within reasonable bounds of human understanding. However, 10% of all the stations seems to defy gravity and have no relation to earth's surface!

Here are my numbers:
Geocoded successfully: 1790 stations
Failed to geocode: 164 stations

Those 164 stations include amazing jewels pointing to a complete lack of standards.
My personal favourites:

NO FRILLS
LAWSON at OVERPASS ( I will be an old man when a geocoder can figure this out)
COMMON AREA ( apparently implies York university and has 7 different ways to spell it)

I've manually geocoded around 100 stations this Saturday and have around 60 left:)
This is a good progress since after I geocode all of them, it will remove the last obstacle for the testing of the faster trip planner:)

1 comment:

Anonymous said...

wise to spring pendants [url=http://www.blingforfun.com]hip come jewelry[/url],[url=http://www.blingforfun.com/pendants/cat_9.html]hip leap pendants[/url],hip bound watches,[url=http://blingforfun.com/belts/cat_18.html]bling bling[/url] ,in skip,[url=http://blingforfun.com/chains/cat_7.html]hip hop chains[/url],wise to caper bling,[url=http://blingforfun.com/chains/cat_7.html]iced out like a light chains[/url],[url=http://www.blingforfun.com/chains/cat_7.html]wholesale chains[/url]
hip skip jewelry