On 13 May 2010 17:23, Robert Scott <li...@humanleg.org.uk> wrote: > Hi all, > > I've been running some countrywide comparisons of the recently released OS > Locator against the streets in OSM, using fuzzy string matching and the > supplied bounding boxes to attempt to match each street in each dataset to > one in the other. It's worked pretty well for most areas I tested. Of the > ~826k named streets in OS Locator, about 424k of them have near perfect > matches in OSM. A few tens of thousands more have what I would call spelling > 'disagreements'. The rest of them have bad or no matches at all. > > I've put a description of the technique up here along with the preliminary > results: > > http://humanleg.org.uk/code/oslmusicalchairs > > The thing I really need is suggestions for getting this data to users in a > way that's practical to work with. It's a CSV currently. > > Thoughts welcome. So are bug reports of where my matching algorithm has > gotten things wrong. >
What about using double metaphone for finding spelling disagreements? Emilie Laffray
_______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
