[OSM-talk] Need help!

Fri Sep 12 21:04:12 BST 2008

I disagree, at least for my approach. Even when people do maintenance,
they download a lot more things than they change. My approach will
also identify "islands" of users whose work are not reviewed by the
"core" community.

But with or without my tool, we badly need a simple log to track the
actions of newbies.

--
I hope it's not the start of a new trend, but I find that I have been
doing more reverts than maintenance lately : A certain user has
changed all the highway junctions in Johannesburg from motorway_link
to trunk.

On Fri, Sep 12, 2008 at 9:42 PM, Shaun McDonald
<shaun at shaunmcdonald.me.uk> wrote:
> Unfortunately this won't quite work as we move from a data entry to a data
> maintenance mode. In the data maintenance mode you will generally be looking
> to update the data as the world changes. This is where this method appears
> to fall down.
>
> Shaun
> On 12 Sep 2008, at 19:28, Nic Roets wrote:
>
>> Even better would be a review system similar to Page Rank for web
>> pages : If user X downloads n objects edited by Y and does not revert
>> it nor mark it as bad, we assume user X has a good opinion of user Y's
>> work.
>>
>> In terms of web pages, it's like user X adding n links on his page to
>> user Y's page.
>>
>> So we create this directed graph with 1 node of each user and the
>> weights on the edges are the number of objects downloaded and not
>> edited or flagged. Then we let a particle is let loose at the root
>> node (SteveC) and let that particle make a random walk through the
>> graph based on the weights. The "score" (or Page Rank) of each user Z
>> is then the probability that the particle is at node Z at time
>> infinite. Then we monitor users with many edits and a low score.
>>
>> For 60,000 users the maths does not require a super computer. Just a
>> good eingen value / vector library.
>>
>> Regards,
>> Nic
>>
>> On Fri, Sep 12, 2008 at 4:40 PM, Peter Miller <peter.miller at itoworld.com>
>> wrote:
>>>
>>> I realise that this particular incident was an accident, however can I
>>> suggest that we add something to be able to check up on a user so as to
>>> get
>>> an idea about them and the intentions.
>>>
>>> General information- this information is to judge how trusted a
>>> contributor
>>> is and if their contributions should be checked:
>>> 1) When did they register?
>>> 2) When did they do their first edit?
>>> 3) On how many distinct days have they made edits?
>>> 4) How many compliments have been made about their work by other users?
>>> 5) How many complaints have been made about their work by other users?
>>>
>>> Then an indication of activity...  How about a table summarising edits
>>> over
>>> the past 10 weeks and the last 10 days:
>>>
>>> 1) How many ways added?
>>> 2) How many ways modified/split/combined?
>>> 3) How many ways deleted?
>>>
>>> Finally a list of recent textual changes (which might be a likely target
>>> for
>>> vandalism); the latest changes to visible textual fields, ie name, ref.
>>> for
>>> example:
>>>
>>> High Street (deleted)
>>> French Street (added)
>>> John Street -> Burt Street (change)
>>> A14 -> blar blar blar (change)
>>> Etc
>>>
>>> Could we also add a monitor to the minutely update automatically for
>>> worrying stuff? What is worrying can be added to over time, howver we
>>> could
>>> start by looking for large changes by new users, the occurrence of 'high
>>> risk' text field values etc.
>>>
>>> Of course it is also be useful to monitor all changes in an area as one
>>> can
>>> with our 'OSM Mapper'product:
>>> (http://wiki.openstreetmap.org/index.php/OSM_Mapper). Using this product
>>> I
>>> have spotted a few new good local mappers appearing in my area and been
>>> able
>>> to check their initial work and help them with tagging etc. It might be
>>> good
>>> to be able to give them recognition for their work (some 'gold star' or
>>> whatever) which might then reassure other users if they start working in
>>> a
>>> different part of the world. I could also flag up a concern first
>>> privately
>>> and then possibly publically if their work was not good.
>>>
>>>
>>>
>>> Regards,
>>>
>>>
>>>
>>>
>>> Peter
>>>
>>>
>>>
>>>
>>>
>>>> -----Original Message-----
>>>> From: talk-bounces at openstreetmap.org [mailto:talk-
>>>> bounces at openstreetmap.org] On Behalf Of Tom Hughes
>>>> Sent: 11 September 2008 15:34
>>>> To: Frederik Ramm
>>>> Cc: osm Talk; Дмитрий Оленёв
>>>> Subject: Re: [OSM-talk] Need help!
>>>>
>>>> Frederik Ramm wrote:
>>>>
>>>>>>>> * : See http://www.openstreetmap.org/stats/data_stats.html
>>>>>
>>>>>>> Heh. Dmitry Oleynov is the most destructive mapper. For two days (8-9
>>>>>>> of  Sept) he has removed 75139 nodes! But created only 888.
>>>>>>
>>>>>> His actions do not appear to be malicious though - in fact he emailed
>>>>>> Steve on Tuesday after he realised what he had done asking for help to
>>>>>> revert the changes.
>>>>>>
>>>>>> That got passed on to me, but I don't any way of reverting stuff, so I
>>>>>> suggested last night that he come here and ask for help.
>>>>>
>>>>> I have a script which I could use to
>>>>>
>>>>> 1. create a list of nodes and ways edited by a certain user in a
>>>>> certain
>>>>> time frame, planet-wide
>>>>>
>>>>> 2. load the history of all these objects from the current database
>>>>>
>>>>> 3. for each object that has not been touched by someone else since,
>>>>> revert it to the state it was before the user touched it.
>>>>>
>>>>> If you want me to do that for changes by Dmitry (and if I may assume
>>>>> that he is happy with this) then just give me a time frame and the
>>>>> username.
>>>>>
>>>>> This procedure has two disadvantages: 1. it will also remove "good"
>>>>> changes by that user in that time frame, and 2. it will NOT touch
>>>>> anything modified by someone else - this means that wherever people
>>>>> have
>>>>> attempted to manually repair the damage, things will remain in the
>>>>> "manually repaired" state and not be reverted.
>>>>
>>>> Dmitry - are you happy for Frederik to do the above to revert your
>>>> edits?
>>>>
>>>> Tom
>>>>
>>>> --
>>>> Tom Hughes (tom at compton.nu)
>>>> http://www.compton.nu/
>>>>
>>>> _______________________________________________
>>>> talk mailing list
>>>> talk at openstreetmap.org
>>>> http://lists.openstreetmap.org/listinfo/talk
>>>
>>>
>>> _______________________________________________
>>> talk mailing list
>>> talk at openstreetmap.org
>>> http://lists.openstreetmap.org/listinfo/talk
>>>
>> _______________________________________________
>> talk mailing list
>> talk at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/talk
>
>