I recently forked this project to make some improvements in the zip code lookups and in street matching. For the latter I utilized the sqlite spellfix library which has much better spelling performance than the metaphones currently in the project. This permits successful matches where there are larger discrepancies in the spellings. I also modified the zip code search to look at all zip codes in the county of the original zip code, rather than just dropping the last two digits. This manages to search the most likely zip codes, and certainly those adjacent to the original zip code in that event that it was changed at some point in the past (this is more useful when working with old addresses). Also, I introduced some thresholds to induce further lookups if poor matches were returned. Please feel free to utilize this fork and contact me if there are errors, or if clarification is needed. The readme contains a longer description of the changes as well as instructions to compile the spellfix library and load the data.
https://github.com/mb16/geocoder
I recently forked this project to make some improvements in the zip code lookups and in street matching. For the latter I utilized the sqlite spellfix library which has much better spelling performance than the metaphones currently in the project. This permits successful matches where there are larger discrepancies in the spellings. I also modified the zip code search to look at all zip codes in the county of the original zip code, rather than just dropping the last two digits. This manages to search the most likely zip codes, and certainly those adjacent to the original zip code in that event that it was changed at some point in the past (this is more useful when working with old addresses). Also, I introduced some thresholds to induce further lookups if poor matches were returned. Please feel free to utilize this fork and contact me if there are errors, or if clarification is needed. The readme contains a longer description of the changes as well as instructions to compile the spellfix library and load the data. https://github.com/mb16/geocoder