I've noticed in the GB.txt dump there seem to be a lot of duplicates, the only difference is that these rows have different feature codes and feature class but all admin codes are identical?
Below is a sample but I have over 7000 rows that are like this.
Code:
geonameid | name | f_class | f_code | country | admin1 | admin2 | admin3 | admin4
---------------------------------------------------------------------------------------------
2657876 | Abbey Dore | P | PPLA3 | GB | ENG | F7 | 00GA001 |
7291410 | Abbey Dore | A | ADM3 | GB | ENG | F7 | 00GA001 |
7296023 | Bolton | A | ADM4 | GB | ENG | C9 | 16UF | 16UF008
11592596 | Bolton | P | PPL | GB | ENG | C9 | 16UF | 16UF008
The problem is how can you build any meaningful autocomplete service if you search for a place and when you return the full concatenated place name with all admin levels the end user is having to select between two identical names?