[OSM-talk] Mechanically Cleaning Up FIXME Tags

[moltonel 3x Combo]

On 26/02/2015, Bryce Nesbitt <bryce2 at obviously.com> wrote:
> *One)* We have a "fixme" system where human mappers are encouraged to pay
> extra attention to particular areas or objects.
>
> *Two) *There is an issue of mapper fatigue: each mapper will look at only
> so many such tags in a lifetime of mapping.
>
> *Three)* The fixme system is not self-cleaning.  Certain conditions result
> in fixme tags that are unlikely to be acted on.  There are some 1.3 million
> open fixme tagged items, more than half from mechanical tagging.
>
> *Four)* In some cases the fixme tags happen to be associated with poor
> quality imports. But this is not universal: some poor data has fixme tags,
> other poor data does not.

+1 on all that, except that 4) is barely relevant. If an import is so
bad that it needs to be undone, I really hope that the presence of a
fixme tag is not the only way to detect said import.

> -------------------------------------------------------------
> How about a two step process:
>
> *Step One ) * People who wish to delete a particular import look through
> the FIXME tagged items, and  propose specific deletions.  For example
> there's a bus stop import that looks to be of bad quality.  If that data is
> removed, the fixme tag will go with it. Problem solved.  *Make a specific
> proposal showing why the fixme tag is needed in order to clean the data.*

Fair enough, but note that the problem being solved is the bad import,
not the distracting fixme tags.

> *Step Two )  *Remaining fixme values with a count above 10000 are
> reviewed.  If they are deemed to add value, or if they come from
> many hand mapping efforts, they stay.  The rest are mechanically trimmed.

The usual "find a frequently-used tag that ought to be deleted an
maybe its associated data fixed" process then. Not really specific to
fixme tags, until you point out a particular fixme value that deserves
the treatment.