[talk-au] poor business listing edits

[Andrew Davidson]

On 10/05/18 17:18, Andrew Harvey wrote:
> These edits are hallmark seo spam, likely all done by the same 
> organisation, following the same instructions. They all have the same 
> traits (new username for each edit they make, named after the company, 
> abuse the changeset comment with spam, never use a primary key to 
> describe what it is they are adding, never use the correct format for 
> phone, never reply to changeset comments, always have spammy 
> description, and sometimes add business which don't have any on the 
> ground presence where they add the node. The organisation(s) behind 
> these systematic edits have had enough time to learn and work with the 
> community but they have shown no interest in doing that.
> 
> Honestly given they don't take changeset comments into account, I think 
> it's a waste of time to try to help them edit better, we've tried and it 
> hasn't changed the way they work with OSM.

On 10/05/18 17:52, Philip Mallis wrote:
 > Agreed. Spent a lot of time removing spam business edits in and around
 > Federation Square in Melbourne. They are fairly obvious spam accounts
 > and I don't see benefit from trying to engage with them.

I've now sampled at least 100+ of these delights. They are currently 
stupidly easy to find with OsmCha, but no doubt that will change in the 
future. What I've found is, if I'm being generous, only 1 in 4 of these 
things are worth trying to save. Highlights I found:

* In the early days they were using OSM accounts under their own names:
https://www.openstreetmap.org/user/sam%20davies/history
https://www.openstreetmap.org/user/Eddie%20hanham

Till they got blocked by the DWG.

* Some of them have tried to create user profile that appear to be "real 
people":
https://www.openstreetmap.org/user/KylieJBridal
https://www.openstreetmap.org/user/TracyAshley
https://www.openstreetmap.org/user/BrandonCrayden81

* Only one of them managed to use a physical tag (yeap, I was genuinely 
surprised)

* The description tags and changesets were almost all "spammy" and in a 
lot of cases cut-and-pasted from the website of the business and just 
cut off at the 255 character limit.

* In a majority of cases I could not verify the physical address from 
the website of the business.

* There were more than a few fake addresses. In some cases the business 
was listed in the Yellow Pages under a different address.

* iD has a helpful address preset that allows them to create plausible 
looking fake addresses by just generating a random street number and 
then selecting the street/suburb from the drop-downs. Which means you 
can't just rely on the node being roughly in the location of the address.

* A handful of them asked for reviews ?!

* Lots of collateral damage in the form of dragging existing nodes, 
using suburb labels, and just randomly changing tags on other things.

* It takes 3-5 minutes to check and fix each one.

* Finally, some of them came back and did some follow-up editing:

This one was probably the closest to not being spam:
https://osmlab.github.io/osm-deep-history/#/node/5621452205

and this one I de-spammed and then they came back and tried to re-spam:
https://osmlab.github.io/osm-deep-history/#/node/5563230416

After wasting many hours on this I've come to the conclusion that a 
zero-tolerance policy is the only real option we have. So I've now 
started to just revert them without bothering to look to closely.

On 10/05/18 10:29, Warin wrote:
 > In reverting, OSM looses the information that these new people made.
 > OSM also looses a potential new mapper, as I doubt they will return
 > following their attempted addition being removed.

If you want to send your time re-doing the work of people who are 
getting paid to do this then knock yourself out. All I ask is that you 
mark them as good in OsmCha, otherwise it will be goodnight Vienna.