Improve vehicle make and model cleaning

[Robinlovelace]

Based on feedback in vehicle-comments.csv, we need to improve the cleaning logic for vehicle makes and models.

Top 10 High Impact Fixes:

1. Expand multi-word make extraction (e.g., Freight Rover, Austin Morris, Austin Morris, etc.)
2. Fix model capitalization for common abbreviations (CBR, RS, SQ, GS, etc.)
3. Handle Mercedes-Benz 'Class' models capitalization (e.g., CLA Class, GLA Class)
4. Standardize/Collapse Vauxhall and Opel makes.
5. Standardize/Collapse MCC and smart makes.
6. Allow 'Generic' as a valid make (currently treated as NA).
7. Refine DAF and Leyland DAF distinction and standardization.
8. Consistent handling of 'Model Missing' and redacted values.
9. Fix specific model capitalization (e.g., bZ4X, LandCruiser).
10. Handle Vespa (Douglas) as a distinct make.

These changes will significantly improve the quality of cleaned vehicle data.

[wengraf]

If I were to order these jobs in terms of priority, it would be 1, 8, 4, 6, 7 and 10, ordered by high to low. The others in any order. If Opel/Vauxhall aren't to be collapsed, at the very least Opels shouldn't be turned into Vauxhalls (even though they are largely the same).

[Robinlovelace]

If I were to order these jobs in terms of priority, it would be 1, 8, 4, 6, 7 and 10, ordered by high to low. The others in any order. If Opel/Vauxhall aren't to be collapsed, at the very least Opels shouldn't be turned into Vauxhalls (even though they are largely the same).

Fair, I will keep them separate.

[wengraf]

I'll definitely have a go with this today or tomorrow.

[Robinlovelace]

Great, thanks @wengraf it seems that @elifclaus is on the case.

[Robinlovelace]

I expect we'll see more results flowing in #298 and it may just be a case of checking the outputs and flagging anything we're missing, it would be great if you could review that PR Ivo.

[PublicHealthDataGeek]

I have found three values in the generic_make_model that do not work correctly with the functions. 'White Knuckle 125' goes to NA with clean_make(), Knuckle 125 with clean_model() and Knuckle 125 with clean_make_model(). Example collision_index are 201232N335812 Similarly, if a vehicle generic make model is 'GENERIC CODE 125' or 'GENERIC TRIGGER' (e.g. collision_index == 2016471604583) but these appear to be just generic make model codes so probably less important.

[wengraf]

I have found three values in the generic_make_model that do not work correctly with the functions. 'White Knuckle 125' goes to NA with clean_make(), Knuckle 125 with clean_model() and Knuckle 125 with clean_make_model(). Example collision_index are 201232N335812 Similarly, if a vehicle generic make model is 'GENERIC CODE 125' or 'GENERIC TRIGGER' (e.g. collision_index == 2016471604583) but these appear to be just generic make model codes so probably less important.

Worth remembering that "Generic" in Make/Model data is always the name of a make, not a generalisation.

[Robinlovelace]

Wow that is confusing! Generic != generic...