[Talk-us-massachusetts] Missing and unnamed streets

[Yury Yatsynovich]

Greetings!
There is definitely no presumption that in any conflict between MAD and OSM
MAD is correct.
I've gone through problem address points (from fuzzy matches) in several
towns in Berkshire and my impression is that MAD is correct in 80-90% of
cases when OSM has some road names (I was looking at GIS maps by local
authorities, and Mapillary images for road signs -- that's why it also
takes me around 1 hour per town).
Another observation -- MAD points with housenumber=0 have much higher
probabilities of being at a wrong location than others.

Here is the spreadsheet with streets that are in the unmatched MAD points
(after the fuzzy match), but not in the boundaries of corresponding towns
(9K in total, median -- 15 streets per town):
https://mega.nz/#!TxlABapS!4MNK5R4UQEkeLjj1pNISufQMNgGs2Vl1YvLGFU4RU3A

As I posted earlier, the link to the files after the exact match (split by
towns): https://mega.nz/#F!ToFnGI6C!jcjnjc3753w4DcfaSLLAJg
Ideally, we would want to address all discrepancies (including those like
"Street" vs "Lane", "Muller's" vs "Mullers", "Hilltop" vs "Hill Top"), yet,
the amount of them is even larger than of those after the fuzzy match. So,
I would suggest posting on MapRoulette first the list of 9K streets from
the fuzzy match and, once those are done (manually, with checking multiple
sources on street names), we can address the discrepancies from the exact
match.

Talking about the stacked points, I would import them (those that pass our
regular checks), converting into 1 point and combining the same fields as
addr:housenumber=1;3;5 etc. The share of addresses assigned to stacked
points in MAD is substantial: most multi-unit buildings have address points
at their centroids rather than at entrances.

-- 
Yury Yatsynovich
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us-massachusetts/attachments/20180815/6746cfad/attachment.html>