[Talk-ko] possible legal problems and next thing to do

cyana pyroshot at hotmail.com
Fri Apr 9 13:15:11 BST 2010


EXPLANATION

'cyana's POIs' mainly gathered from freely available sources
such as school directory, hospital directory, etc...
I assured when someone asked me personally that POIs may not
have legal probelm. true but need to be explained.
I didn't mentioned it properly or wrote in POI's 'source' tag clearly.
Because it was collected from so many sources and some sources
doesn't available anymore or cannot find it again.
Even I didn't know about massive POI upload info wiki pages exists on
OSM wiki!

Here's what i said to J. Stein before
(also see in http://wiki.openstreetmap.org/wiki/Koreaimport09)

BEGIN
-snip-
And then get the lat/lon pair from yahoo's open API POI service
(yes it was really nice gift, free service in korea).
Yahoo korea provides service which can return lat/lon pair from
the given unique name. It covers most of massive POIs uploaded
last year by me. Some of POIs from government such like Seoul
bus POI database, metro subway databases. all databases was public.
But I didn't set proper OSM tags for those POIs.
I just set that POIs was coming from Yahoo Korea using
some script as you know.
-snip-
END

Add some details for my word:

I didn't get all of POI's lat/lon pairs from yahoo.
Because Yahoo Korea POI query service only provides a handfull
hundreds POIs per day per unique registered user.
And many POI info didn't get in that way because name ambiguity
is very common in Korea.

ie. '속초횟집 (Sokcho Fish Restaurant)' is a popular name and
a lot of same entries with different lat/lon in Yahoo Korea.

Case #1: Here's what I got locations from Yahoo Korea:

* unique name: like administration district with known address.
note: It was already provided by Geo Names Server or free Korea
Traffic Information Service database, even in old U.S military map.
So I assume it as public.

* other unique names: like railway stations. I got name database,
and locate it with yahoo. It would makes some legal problem.
How many of them? provincial offices, regional government offices, etc.
I didn't count yet. I have to inspect it carefully, get/replace
those POIs from other sources. Because Yahoo's service restriction
and ambiguity problem, I didn't get it much from Yahoo.

Case #2: Here's what I get location *reference* from Yahoo Korea which
POIs that has lat/lon pair already:

* approx. 30000 POIs from my personal collection (no source, no license
info, no copyright notice)
* approx. 80000 POIs from Korea National Traffic Information Service
database
(a.k.a. NTIC, public. used it when I was upload massive road data).

-> I did randomly sampled those POI's location infos using
Yahoo Korea POI API and compared it because many POIs
what I got have bad projection problem. korea uses
various projection method: KATEC, TM, Tokyo.
Last year, our government decide to use WGS-84 as a standard
but problems remains for old, never-looked-up-again POIs.

If it matched with-in some geographical distance
(usually 50m~100m. I'm not sure, I've changed that variable time to time),
I accept it as a valid POI because Yahoo & Naver service also
has same problems in 'location accuracy'.

Actually, I didn't fully inspect all POIs.
IMHO, It would not a problem at all, I didn't replace their location
with Yahoo's and it's not matched with their lat/lon exactly.
In this case, I will remove all 'source=yahoo' tags completely.
It was tagged even it wasn't obtain the lat/lon pairs from Yahoo.

fact: Some random samples was verified using Yahoo occationally,
but tagged as source=yahoo.
maybe replace 'source=yahoo' tag to 'verified-some-with=yahoo' tag. sigh.

MT-INFO

I've collect many mountain peak, valley, hill, cultural/touristic
spot POIs from various sources several monthes ago.
It is heavily duplicated because it was converted from various
GPX which is uploaded to trekker's society. yes, it's free.
It has some problem in location accuracy,
or by bad projection method/conversion errors/human errors,
or even they just press the GPSr button to record POI accidentally.

ie. 17 'Mt. Soeback' POI collected from NTIC, my GPS data,
and other sources located different place when I saw in the map.
6 of them nearly peaked in right place but not exactly
positioned highest point in my topo map overlay.
I've removed 11 completely out-of-bound POIs,
and triangulate 6 POIs for 2 POI.
And then move 2 POIs to near highest position.
It takes for a month for refining only peaks and hills.

1. create custom topo map thas has elevation grid,
2. convert data into common format,
3. remove, triangulation, move to nearist peak.
4. find affordable parameters. repeat step 3
5. verify with my eyes.

1 to 3 is done by my C++ program.
I made a lot of small, temporal mapping job programs for myself.
However, that data will be called 'mt-info'.
It shows more than 92% accuracy by random sampling.
Why am I upload only 92% accuray POIs?
because that POIs mainly created by hand of man.
Like any other human made POIs, it contains many errors.
Verifying all POIs are just humanly impossible...

NEXT JOB TO DO

I promised that remove all duplicated POIs few days ago. How will I do:

0. highway=bus_stop: remove & concatenate. already done today.
1. collecting POIs in South Korea which I(cyana) upload last year, stuff
with 'mt-info'
2. calculate relative geographical location for each POIs from center of
South Korea.
3. remove POIs in same location (condition: same name, less tagged,
using priority)
note: this map
(http://matt.dev.openstreetmap.org/dupe_nodes/?zoom=7&lat=36.30734&lon=127.4773)

shows a lot of duplicated POIs. but in korea, it is really populated in
same location,
I'm sure who visit urban area of South Korea knows what I mean!
4. re-translate, re-romanize tags like name:en, name:ko, name:ko_rm.
5. upload changes

It will takes for a month... I'm doing this job in free time,
so the schedule would be vary at each step.

After finish those job, i will remove all POIs directly
get lat/lon pairs from Yahoo Korea before someone modify it.
I don't worry about delay. Korea mappers didn't modified it much.
And they are very small group in OSM society. :)

As always saying, any suggestions welcome.

P.S. I'm not good at english, such difficulty makes me
shy to interact with other mappers.
I respect OSM's policy, so, don't want to any harmful
job to OSM, my precious...





More information about the Talk-ko mailing list