First result doesn't match string name

[vinxavier]

I am searching for a street called "Francisco Lisboa", it returns me some results, but the first result doesn't match my string, just the second.

Maybe could be nice use some score for see how strings are similar to show first the higher scores strings.

As the image, the first result shows the street "Antonio Francisco Lisboa", but the street that I want is "Francisco Lisboa".

[lonvia]

Actually, neither of these results matches your request completely because both have a 'Rua' in front. Nominatim doesn't really know that 'Rua' is unimportant for you, while 'Antonio' isn't. This kind of distinction requires quite a bit of real-world knowledge and knowledge about how addresses and street names are formulated in the different parts of the world.

[vinxavier]

Kind, but if you see the string for the street have quite a difference. There are algorithms that can calculate how similar a string is. Like the Jaccard Index. Antonio is a huge difference in the query result. It's clear more far from my search "rua antonio ..." then just "rua... " in the front.

https://github.com/ecto/jaccard/blob/master/jaccard.js

[mtmail]

Autovía del Nordeste is different from AV Nordeste by eight characters and it different writing of the same street. Carrer d'Alfons XII is different from Carrer d'Alfons XIII by one character and it's different streets (both exist in Barcelona).

Applying a generic algorithms like Levenshtein distance/LCS/jaccard to compare user supplied input (query string) against results (OSM data) isn't enough.

[lonvia]

Closing in favour of the longer discussion in #679.