geocoding improvements

[cole-brokamp]

• range geocoding improvements #37
  • add match method of tiger_street versus tiger_range #28
  • use s2_interpolate to return interpolated point along line instead of centroid
  • using a street offset to break ties for street ranges that border on census geographies
  • offer option to return s2 cell that is a common parent ancestor of all tiger line ranges
• search "nearby" zipcodes for addresses without matches in exactly matched zipcode addresses #26
• better address tagging in special cases (interstate MM, terrace, etc) #27
• provide option to return numerically nearest street range on matched street instead of the centroid of all the street ranges #29

[erikarasnick]

2400 block of Colerain Ave was matched to the entire length of Colerain Ave. Upon further inspection, I don't think the 2400 block exists... the street is not continuous and jumps from 2100s to 2600s. When our desired outcome is a matched street range geography (as in xx_address) would we want a case like this to return missing? Or the nearest street range?

[cole-brokamp]

Returning the numerically nearest street range would be a good option to provide instead of the union when the street is matched but the ranges do not contain the street number.

edit: tracking this specific change in #29

[erikarasnick]

I noticed when using addr_match_tiger_street_ranges some addrs are not getting matched even when the street does exist in the TIGER street ranges.

e.g.,

addr	TIGER street name
2250 Westwood Northern Boulevard	Westwood Northern Blvd
1050 William H Taft Road	William Howard Taft Rd
2650 Richwill Court	Richwill Ct

The street suffixes are abbreviated in TIGER and not addr, but this doesn't cause a problem for lots of other addresses. Could this be related to the "number of edits" matching?

I also don't see any interstates in the TIGER street ranges. Are those getting filtered out?

[cole-brokamp]

Compare the address tags in those two cases. Maybe getting parsed incorrectly?

[erikarasnick]

yes, so this is related to #27. will track there.

except the Taft one is just the difference in "H" vs "Howard"

[cole-brokamp]

now that we have this broken up into more modular issues, closing tracking here.