Hello all, I've made a few updates to the matching algorithm in musical chairs [1]. First of all, it now checks the fields name, name:en, name:cy, name:gd and alt_name for the best match. Along with the normalization of accented characters, this now allows it to cope with wales a lot better [2]. So you don't have to stick the Ordnance Survey name for a street as the primary name just to get it to show up as a match.
On top of that the levenshtein comparison now only gives half weighting to a spacing edit, which allows us to differentiate between differently spelt names and differently spaced names (which are often borderline/debatable anyway - there are plenty of streets with a name spaced one way on a sign at one end and spaced differently on a sign at the other). This is all made possible by writing a custom c postgres module to make it fast enough to be feasible. Thanks to TomH for trusting me enough to insert it ;) robert. [1] http://ris.dev.openstreetmap.org/oslmusicalchairs [2] http://ris.dev.openstreetmap.org/oslmusicalchairs/map?zoom=8&lat=52.46286&lon=-3.52652&layers=B0TF&view_mode=recentrelevantupdate _______________________________________________ Talk-GB mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-gb
