I replied on the list and posted a veto after a quick check and missing roads reported are only at most highway=track or highway=service. Without local knowledge remote mappers are likely to mess up highway classification more.
I am directing in the reply to this post for further discussion.
Stephan, thank you for your thoughtful reply and for starting this thread in the preferred channel. We are glad to know of this forum as an addition to the OSM Mailing lists since our policy is to always wait for community feedback before beginning our projects. We have removed the Thailand Missing Roads challenge for now and will await your further questions and inputs before we begin work. --CourtneyCW
I am Marjan from TomTom. We are aware that in recent years, you’ve had some bad experiences with errors introduced by organised editing in Thailand (specifically road alignment and categorization issues). We therefore understand that challenges like our Missing Roads project are met with some skepticism. Our goal is to further improve the map in cooperation with the local community. Therefore, to meet these concerns, here is some additional information:
First of all, I would like to ask: where do these suggestions of missing roads come from? Is it based on TomTom routing data or some imagery-based algorithm?
Like @stephankn mentioned, our main concern is that in most cases highway classifications can’t be simply guessed without a ground survey or local knowledge. Adding road geometries is easy and can be done by anyone including local mappers.
I hope the additional feedback above answers your questions and concerns. If no one has any objections, could our team move forward with the project? If yes, we would be checking in regularly to see if you have additional advice or questions.
I appreciate that you answered back. But you just gave some very unspecific answers, which do not address the concerns we raised.
First: You created a MapRoulette challenge. This is open to anyone. And based on experience very often they are picked up by people not familiar with the local situation at all and just wanting to support OSM.
So your point of pointing YOUR editors to something does not help. You have no control about the edits happening via MapRoulette.
Your project does not give any hint for mappers on how to do proper highway tagging.
Referring to some general best practices does not help at all here.
What are your planned quality control steps of the contributions? Also considering that these edits happen on live data. So any correction would happen after damage was already done.
Please re-read our initial posts. We checked roads marked by your algorithm as missing.
Youhave a nearly 100% rate of “missing” roads impossible to tag correctly by remote mappers unfamiliar with the situation on the ground.
Why is it so important for you to have roads added which are at most highway=track or maybe service? They play only a minor role in the road network at all. Is it for you to satisfy some metric created by an algorithm pointing out to “unmapped” roads?
We could certainly live with a tiny percentage of mis-classified roads. But this is aiming at a failure rate of maybe 80% which is unacceptable, giving an optimistic guess that you might be able to deduct the classification correctly from context on some highways.
We hear your concerns and will not start this project in Thailand. Let me reassure you that we haven’t made any edits, and we would not have made any edits without your collaboration. The sincere goal of our notification in the mailing list was to propose this challenge and see if any questions arose. Our goal is to improve the map and help the editing community. You are asking good questions, and we understand your concern.
Though we will not work on this project, we do still believe we have expertise and tools that might be useful to you. Should you wish to brainstorm other ways that we might be able to support your local efforts, please reach out.
Do let us know if you see any ways in which we could help you further improve the OSM map in Thailand.
I had hoped that others would chime in with ideas on what their specific topics would be where we could benefit from tooling support.
I have some ideas in mind, but do not want to push the discussion already in a specific direction. I have done tool support in Thailand before on various areas. Also running checks on data completion and have experimented with machine learning.
And by running the Osmose instances i know that a lot of things can break and that even more corner cases exist which make tool support in terms of data verification difficult.
So to start of the brainstorming:
What kind of tasks would you see done?
Completion of road network? How do you define “road-network”? What road classoification is for you part of this? Major highways, minor highways, local roads and agricultural tracks or even hiking paths?
Against what metric would you measure the completion?
Fixing of geometry problems on the data? What kind? Unconnected highways? Invalid/Overlapping areas (like buildings)?
Routing problems like unreachable roads/islands/traps?
Tagging inconsistencies like sporadic classification change of networks or inconsistent attributes?
What specific tool support do you see on the areas you would like to improve?
Thank you for your detailed response. We really appreciate your willingness to brainstorm with us. Feedback from someone who has experience with tooling support in Thailand is very valuable to us.
To detect potential OSM data issues that we can fix, we are using public instances of OSM QC tools such as Atlas and Osmose in addition to internal validation. In Thailand, we have detected the following types of issues:
not connected highways (cases where the end of the highway is not connected to another highway),
highways with impossible angles,
road name gaps (cases where the name tag of a way is missing or different from the name tag of the ways connected to it).
Our idea would be to generate MapRoulette tasks to fix the detected issues. Members of the Thailand community can take part in these challenges if they are interested.
We could share this data in about a week’s time if you approve, but we will await your feedback before we proceed.