API Docs: https://dev.socrata.com/foundry/data.baltimorecity.gov/m8g9-abgb
Geocoding
Is any information about the geocoding process for this dataset available? There appear to be a number of valid human readable locations ("incidentLocation" column) that do not appear to be geocoded to a valid location ("location" column) and some invalid locations geocoded to a valid geographical location.
A list with entries that could not be assigned a neighborhood from the past day can be found at https://www.neighborhoodcallwatch.com/Uncategorized. 50 of the 118 entries do not have a location (either "(,)" or "(0.000000,0.000000)") and the remainder either have "100" or similar for the "incidentLocation" and a "location" that does not appear to correspond to a Baltimore City neighborhood (e.g., "2600 ROYAL OAK AV").
Sample problematic entries:
"MONDAWMIN MALL", a recognizable place, has been geocoded to "(0.000000,0.000000)" for call "P163321463" and "(39.318644,-76.654012)" for call "P163321732". "P163321463" appears to have occurred before "P163321732".
"1300 DOCK ST" was geocoded to "(,)" for calls "P163321083" and "P163321131".
"HQE" was geocoded to "(0.000000,0.000000)" for call "P163321606".
"100" was geocoded to "(39.3810499,-76.6572571)", which appears to be outside of city limits, for a number of calls, among them, "P163320024". Does 100 have a special meaning?
Description
Each call entry appears to have a description to summarize the issue ("description" column). However, the number of unique descriptions appear to be in the thousands when counted by year and close to one thousand when counted by month with some duplication/typos.
Is any information available with regard to the data normalization/a list of terms that should be expected for comparisons across time/areas?
Thanks!