OpenStreetMap
Bot idea: Fixing invalid capitalization of primary tags
Unviáu por bkowshik el 9 March 2017 en EnglishListed on the Map Features Wiki are 26 primary features that include highway, natural, building, etc. I ran a tile-reduce script looking for invalid capitalization in the 26 primary feature tags. Examples of invalid capitalization are:
Highwayinstead ofhighwayLANDUSEinstead oflanduse
I got back 3,186 features on OpenStreetMap with invalid capitalization in the primary feature tag. Ex: way/476771384
Such features are usually reported on QA tools like keepright and fixed by the community after some time, but a more constructive approach would be to automatically correct the issue and give the mapper feedback on what was fixed. This positive feedback cycle can help new mappers learn as well as reduce the burden of fixing trivial map issues on others.
Can we build a simple build a bot to do this:
- Continuously monitor feature changes on OpenStreetMap for invalid capitalization in the primary tag
- Automatically correct the invalid capitalization in a new changeset using the chageset create API Ex: Modify
Buildingtobuilding - Let the user know about the modification by posting a changeset discussion on the user’s changeset with a description and the ID of the automatically corrected changeset.
The OSM policy on automated edits does not seem as detailed as the one on Wikipedia which has a well documented system of proposing and operating bots. Are there well documented examples of active bots that can be used as a guideline for proposing a new one?
Please let me know what you feel about this idea and what is the best way to move forward.
Discussion
Comentariu de Zverik el 9 de March de 2017 a les 12:26
I am okay with this. I’d recomment creating a wiki page detailing this automated edit, publishing the bot code on github and after that posting to tagging@ mailing list.
Comentariu de Glassman el 9 de March de 2017 a les 16:27
Another alternative might be Maproulette which means the edit gets a human to verify the object.
Comentariu de BushmanK el 9 de March de 2017 a les 17:57
@Glassman, Keepright does basically the same, doesn’t it?
Comentariu de Glassman el 9 de March de 2017 a les 20:20
@BushmanK - I haven’t spent much time with Keepright - but I suspect you are probably correct. I thought of Maproulette because I recently fixed a number of buildings with self intersecting ways using Maproulette.
Clifford
Comentariu de CloCkWeRX el 11 de March de 2017 a les 08:58
Would also be worth adding such a feature to ID to warn, similar to untagged ways - the less opportunity to accidentally key in invalid data, the less has to be cleaned up by tools later.
+1 in general to the idea; but I’d be interested to know what exceptions/edge cases exist in the data.
Comentariu de amapanda ᚛ᚐᚋᚐᚅᚇᚐ᚜ 🏳️⚧️ el 15 de March de 2017 a les 16:21
At State of the Map 2016 Frederik Ramm gave a talk about mechanical edits. (video here). You should watch it, it’s very relevant to what you’re talking about. He specifically gives examples of typos in common tags like
building. He’s not in favour of automated edits like you suggest, claiming the strength of OSM is having humans looking at things and fixing them, and mechanical edits are not like that.I’d suggest a MapRoulette task to fix up these things if you’d like. You will then have people looking at the area around the problem and can probably find a better fix than what the bot would do.
I suppose one reason Wikipedia has more bots is because Wikipedia is edited with a plain text box, so it’s easy to make typos or simple mistakes. But OSM has software for editing, so people often use presets, rather than having to type in
building=yesall the time. This probably reduces the error rate of mappers.Me personally, I wouldn’t be opposed to mass changing it. However, as you point out, it’s not a lot of instances of this problem, so it wouldn’t be too hard to just use MapRoulette.