Re: [OSM-talk] i18n-rich areas on the map
Emilie Laffray emilie.laffray at gmail.com writes: For example the Imperial Palace in Tokyo would have name:en=Imperial Palace name:jp at Romaji=koukyo name:jp=?? However, I do believe that translitteration is worthy of appearing in name:en when none exists. I agree that when no real English translation exists, then the Romaji version of the Japanese name should be shown instead to English-speaking users. However that's something the map renderer should do; the program which makes the displayed map should know to fall back on name:j...@romaji if there is no name:en available. There's no need to tag it as English just to make it be displayed to English-speaking users; tag it correctly and let the renderer do the right thing. Yes putting it in a different alphabet is not the same, but it can be a starting point until someone is filling the blank with a proper translation hence the two steps: translitteration and a dedicated translation website. Yes. Which is why putting Japanese text (even if it is in the Latin alphabet) into name:en is not a good idea, because it makes it harder to see which things really do need a name:en translation added. I am to some extent a bit annoyed to see things like name = name in native language (English translation) in the OSM files. Agreed. Once multilingual map rendering becomes common, we can expect to see these cleaned up pretty quickly. -- Ed Avis e...@waniasset.com ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Ed Avis wrote: This is not really name:en, more like name:j...@romaji. For example the Imperial Palace in Tokyo would have name:en=Imperial Palace name:j...@romaji=koukyo name:jp=?? Similar considerations apply to countries with more than one alphabet, for example I would expect to see name:en=Belgrade name:s...@cyrillic=??? name:s...@latn=beograd Putting something into a different alphabet is not the same as translating it to a different language, and putting Japanese into a Latin orthography is not the same as translating it to English. So I would suggest adding the Romaji strings if they are needed, but tagging them appropriately and not as name:en. Thank you for this comment and yes, I am quite aware of the distinction for the Japanese language. However, I do believe that translitteration is worthy of appearing in name:en when none exists. I am taking the opposite approach that you are mentionning in this case. In all cases, you are starting in English to go towards the other language. Yes putting it in a different alphabet is not the same, but it can be a starting point until someone is filling the blank with a proper translation hence the two steps: translitteration and a dedicated translation website. However, you have rightly pointed how multiple writings could be used. Maybe a name:Latn would be better in this case or something indicating the language and the destination alphabet. This is an open mail and an open discussion which I believe is worth having. I am to some extent a bit annoyed to see things like name = name in native language (English translation) in the OSM files. I believe that we should keep name:en and name:jp clearly separated. Having fully localized maps for people of those countries would be better. Now, I can see some objections as you being the foreigner you won't be able to read, but those people in those countries won't contribute if they don't see their language displayed in their countries. As the discussion is showing, there are some efforts to have dynamic text layers which I believe is important hence the translitteration effort I am proposing. Emilie Laffray signature.asc Description: OpenPGP digital signature ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Emilie, If you are going to add machine-created romaji transliterations, then I strongly suggest that you put them into name:jp_rm (as given in the Japanese mappers' specification above) and do not put them into name:en. See: http://wiki.openstreetmap.org/wiki/Japan_tagging#Names Using name:jp_rm will be much clearer and consistent than using name:en. It will also be better for using in multi-lingual maps, because the software displaying the map can then be (more) certain which language is stored in each tag and display the correct ones. Automatically creating loads of new name:en tags containing romaji instead of English will increase confusion. I personally think that name:en should only be used for real English translations (although the Japanese mappers' spec. doesn't say that - it allows for transliterations or translations!). Another reason for putting your machine transliterations into name:jp_rm would be to avoid the situation where you add a romaji transliteration into the name:en tag when there is actually a legitimate English translation that should be in there. I can't think of a way for your software to know whether there is a legitimate English translation or not, given that it requires local knowledge (at least some of the time)? It would be much better to have the name:en blank in cases where there is a legitimate English translation (and the translation has not been entered yet!). As many of the Japanese edits are going to be entered by Japanese natives who may not know the legitimate English translations, I'd guess that there are going to be quite a lot of blank name:en tags that should have an English translation not a romaji transliteration, so 'blank' can't be automatically interpreted as 'needs romaji transliteration'. Having just re-read your posting, I'm actually not so sure what you are proposing - you wrote I believe that we should keep name:en and name:jp clearly separated. but than you also wrote I do believe that translitteration is worthy of appearing in name:en when none exists. Hmmm! Cheers, Woll (mapper in Japan) Emilie Laffray wrote: Ed Avis wrote: This is not really name:en, more like name:j...@romaji. For example the Imperial Palace in Tokyo would have name:en=Imperial Palace name:j...@romaji=koukyo name:jp=?? Similar considerations apply to countries with more than one alphabet, for example I would expect to see name:en=Belgrade name:s...@cyrillic=??? name:s...@latn=beograd Putting something into a different alphabet is not the same as translating it to a different language, and putting Japanese into a Latin orthography is not the same as translating it to English. So I would suggest adding the Romaji strings if they are needed, but tagging them appropriately and not as name:en. Thank you for this comment and yes, I am quite aware of the distinction for the Japanese language. However, I do believe that translitteration is worthy of appearing in name:en when none exists. I am taking the opposite approach that you are mentionning in this case. In all cases, you are starting in English to go towards the other language. Yes putting it in a different alphabet is not the same, but it can be a starting point until someone is filling the blank with a proper translation hence the two steps: translitteration and a dedicated translation website. However, you have rightly pointed how multiple writings could be used. Maybe a name:Latn would be better in this case or something indicating the language and the destination alphabet. This is an open mail and an open discussion which I believe is worth having. I am to some extent a bit annoyed to see things like name = name in native language (English translation) in the OSM files. I believe that we should keep name:en and name:jp clearly separated. Having fully localized maps for people of those countries would be better. Now, I can see some objections as you being the foreigner you won't be able to read, but those people in those countries won't contribute if they don't see their language displayed in their countries. As the discussion is showing, there are some efforts to have dynamic text layers which I believe is important hence the translitteration effort I am proposing. Emilie Laffray -- next part -- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 195 bytes Desc: OpenPGP digital signature Url : http://lists.openstreetmap.org/pipermail/talk/attachments/20090729/77e39b82/attachment-0001.pgp -- Message: 3 Date: Wed, 29 Jul 2009 20:16:32 +0200 From: Martin Koppenhoefer dieterdre...@gmail.com Subject: [OSM-talk] definition of the main highway-tag To: osm talk@openstreetmap.org Message-ID: 7acdb3650907291116g4f87609fpbc33a23104a09...@mail.gmail.com
Re: [OSM-talk] i18n-rich areas on the map
Very good point, and feature suggestion. I've started a wiki page on this project. Please contribute! http://wiki.openstreetmap.org/wiki/Map_Translation_Interface From: Ed Avis e...@waniasset.com To: Subhodip Biswas subhodipbis...@gmail.com; Mikel Maron mikel_ma...@yahoo.com Cc: Ævar Arnfjörð Bjarmason ava...@gmail.com; Stefan Baebler stefan.baeb...@gmail.com; talk@openstreetmap.org; David Sasaki osopec...@gmail.com Sent: Monday, July 27, 2009 4:24:00 AM Subject: RE: [OSM-talk] i18n-rich areas on the map Not all map features need translating in the same way. For now, let's assume that only the 'name' tag will be translated, since that is about the only place where freeform natural language is used. (Possibly 'note', 'FIXME' and other things like parking restrictions could also contain natural language, but they are not normally rendered on maps.) Now, there are different kinds of names. There are those which are descriptive, such as 'Town hall' or 'London Zoo', which can always be translated. Then there are names which are purely proper names, such as 'Rome' or 'Brazil'. These may or may not have equivalents in different languages. Many names are in between these two poles, such as Paris's Arc de Triomphe, which may be rendered in English as the Triumphal Arch (taking it as a description to be translated), or kept in the original French (taking it as a proper name). Finally, some things like street names are rarely translated, even if they carry a meaning. Near to me in London is Bread Street, but I would hardly expect to see it on a Spanish tourist guide as 'Calle del Pan'. So I would suggest that any translation interface should accommodate these different levels of translatability. One way might be to record two different kind of translations: one being an equivalent name for a well-known place, and the other kind a translation of the meaning of a string. The web interface would have a tickbox to distinguish the two kinds. -- Ed Avis e...@waniasset.com __ This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email __ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Mikel Maron wrote: Very good point, and feature suggestion. I've started a wiki page on this project. Please contribute! http://wiki.openstreetmap.org/wiki/Map_Translation_Interface Hello, one of the thing that I am working on (at least initially on design) is a translation website, as I strongly believe that the OSM data should be translated as much as it can. I was initially planning to add only data like towns, countries, counties etc.. but I suspect with the help of this page, we can build something that works better and faster. As part of the translation process, I was planning to use translitteration software from Japanese to Romaji to fill potentially missing name:en tags in Japan. Similarly programs exist in Hangul and Chinese. Emilie Laffray signature.asc Description: OpenPGP digital signature ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 6:12 PM, Ævar Arnfjörð Bjarmasonava...@gmail.com wrote: On Wed, Jul 22, 2009 at 4:37 PM, Ed Avise...@waniasset.com wrote: Ævar Arnfjörð Bjarmason avarab at gmail.com writes: http://cassini.toolserver.org/browse-fr.html http://cassini.toolserver.org/browse-nl.html The rest are now up at http://cassini.toolserver.org/tile-browse/ And here's a test wiki to play with: http://u.nix.is/wiki/index.php/Maptest See this maps-l posting: http://lists.wikimedia.org/pipermail/maps-l/2009-July/000158.html ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Very nice! What i miss is getting a bigger map. If there is a thumbnail map embedded into an article people would normally click it in order to see the map in better detail (normally on a bigger map), not to navigate in a thumbnail sized slippy map. There needs to be some intuitive way to switch from the embedded (slippy or static) map to a full screen (or so) slippy map (can be lightbox javascript, so that user doesn't leave the page). If user doesn't have javascript (or has it disabled) clicking on a map should result in a bigger map (with higher zoom to show roughly the same area). Stefan On Mon, Jul 27, 2009 at 10:47 PM, Ævar Arnfjörð Bjarmasonava...@gmail.com wrote: On Wed, Jul 22, 2009 at 6:12 PM, Ævar Arnfjörð Bjarmasonava...@gmail.com wrote: On Wed, Jul 22, 2009 at 4:37 PM, Ed Avise...@waniasset.com wrote: Ævar Arnfjörð Bjarmason avarab at gmail.com writes: http://cassini.toolserver.org/browse-fr.html http://cassini.toolserver.org/browse-nl.html The rest are now up at http://cassini.toolserver.org/tile-browse/ And here's a test wiki to play with: http://u.nix.is/wiki/index.php/Maptest See this maps-l posting: http://lists.wikimedia.org/pipermail/maps-l/2009-July/000158.html ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Hi! On Sat, Jul 25, 2009 at 8:22 AM, Mikel Maronmikel_ma...@yahoo.com wrote: There's a huge community of translators on the web. OpenStreetMap can work with them directly. Wikipedia, Global Voices, Meedan, and many more working on content. Ubuntu, Fedora, OpenOffice, and many many more working on software. Just a few weeks before SotM, the Open Translation Tools conference was held in Amsterdam :) (http://www.aspirationtech.org/events/opentranslation/2009) Translators are used to a certain kind of interface, typically lists of phrases to translate. We can build an interface more welcoming for translators, and tap into this huge community. They want to simply translate, not add new data to the map .. and as OSM grows, we're going to need to offer multiple interfaces to appeal to different types of contributors What I envision is a simple app, with a slippy map. Zoom/pan to the area of interest, select the types of things you want to translate, and submit. The app generates a list, with the name of the object, the localised names in the languages you're interested in, and a small inset static map for reference. Add your translations, submit. The app gets authorization through OAuth, and submits your changeset. Anyone interested in helping build this kind of app? Any other ideas? Let's do it! One such good translation management system is Transifex.It provides most of what we require here.For more details refer to http://transifex.org/wiki/About. However,Is it possible to have an app with slippy map deployment that does the following: 1.Click on the country or state(in case of multilingual country like India) 2.Get routed to the proper translation interface based on coordinates.(For example on clicking on New Delhi leads the translator to Hindi interface.) 3.Translate/manage/edit based on Transifex. Just a thought . -- Regards Subhodip Biswas GPG key : FAEA34AB Server : pgp.mit.edu http://subhodipbiswas.wordpress.com http:/www.fedoraproject.org/wiki/SubhodipBiswas ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
From: Shaun McDonald sh...@shaunmcdonald.me.uk On 23 Jul 2009, at 09:17, Ævar Arnfjörð Bjarmason wrote: On Wed, Jul 22, 2009 at 11:34 PM, 80n80n...@gmail.com wrote: ti...@home uses this technique of a basemap with no captions and a text layer for lowzoom tiles (z1 through z6) . Currently they are combined on the server to create a composite image using some GD library I think. This seems to works very well and avoids duplicating the most compute intensive part of generating a map tile. Yeah it would be neat to render a captionless layer in mapnik and then overlay text on it. Can mapnik support that? I'd have thought adding any sort of text to the map might have implications for the rendering itself. The problem is being able to do the text collision avoidance stuff against non-text features, hence why it's not been done before. How problematic would this be really? The captionless layer could simply orient against the default text, and the localised renderings just did the best they could against that. If text wasn't placed perfectly, well it's less than ideal, but at least we'd have localised tiles! Could this work? I'm simply trying to find a way to localise tiles on osm.org, in a resourceful way! Seems like many people are positive on this. Are there any other ideas for how to set up a localised tile infrastructure? Mikel___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
From: Ævar Arnfjörð Bjarmason ava...@gmail.com On Thu, Jul 23, 2009 at 4:58 AM, Stefan Baeblerstefan.baeb...@gmail.com wrote: Are there any further plans to translate the maps based on wikipedia data? Would that various names be better imported into main OSM db or just into the DB for rendering? Will this translation be done based only on language links in existing articles or also based on lists of exonyms [1], which would cover also places without wikipedia articles? I hadn't planned on doing any sort of automagic OSM - Wikipedia integration for getting i18n names, no. But if you're interested in working on it that's great. It's just a regular open source project a few of us are working on as hobby, so any current plans are just the union of the ideas people are currently interested in working on. There's a huge community of translators on the web. OpenStreetMap can work with them directly. Wikipedia, Global Voices, Meedan, and many more working on content. Ubuntu, Fedora, OpenOffice, and many many more working on software. Just a few weeks before SotM, the Open Translation Tools conference was held in Amsterdam :) (http://www.aspirationtech.org/events/opentranslation/2009) Translators are used to a certain kind of interface, typically lists of phrases to translate. We can build an interface more welcoming for translators, and tap into this huge community. They want to simply translate, not add new data to the map .. and as OSM grows, we're going to need to offer multiple interfaces to appeal to different types of contributors What I envision is a simple app, with a slippy map. Zoom/pan to the area of interest, select the types of things you want to translate, and submit. The app generates a list, with the name of the object, the localised names in the languages you're interested in, and a small inset static map for reference. Add your translations, submit. The app gets authorization through OAuth, and submits your changeset. Anyone interested in helping build this kind of app? Any other ideas? Let's do it! -Mikel ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 11:34 PM, 80n80n...@gmail.com wrote: ti...@home uses this technique of a basemap with no captions and a text layer for lowzoom tiles (z1 through z6) . Currently they are combined on the server to create a composite image using some GD library I think. This seems to works very well and avoids duplicating the most compute intensive part of generating a map tile. Yeah it would be neat to render a captionless layer in mapnik and then overlay text on it. Can mapnik support that? I'd have thought adding any sort of text to the map might have implications for the rendering itself. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Thu, Jul 23, 2009 at 4:58 AM, Stefan Baeblerstefan.baeb...@gmail.com wrote: Are there any further plans to translate the maps based on wikipedia data? Would that various names be better imported into main OSM db or just into the DB for rendering? Will this translation be done based only on language links in existing articles or also based on lists of exonyms [1], which would cover also places without wikipedia articles? I hadn't planned on doing any sort of automagic OSM - Wikipedia integration for getting i18n names, no. But if you're interested in working on it that's great. It's just a regular open source project a few of us are working on as hobby, so any current plans are just the union of the ideas people are currently interested in working on. But personally I'd like to do as little munging (preferably none at all) of OSM data as possible. If someone wants to alter the OSM data in some way (e.g. add exonyms) I'd rather devise ways to funnel that data from Wikipedia (whether automatically or by sending users over to edit the map) into OSM so that everyone can use it from there. Can we help in some way? I'm maintaining a bug list here: http://www.mediawiki.org/wiki/Extension:SlippyMap#Development But that's heavily biased towards what I'm interested in working on (since I'm practically the only one filing bugs), what I'm working towards is basically this: http://techblog.wikimedia.org/2009/04/openstreetmap-maps-will-be-added-to-wikimedia-projects/ But there are all sorts of interesting things that can be done with OSM maps on Wikipedia that need to be implemented eventually. Everything from mundane server setup tasks to harvesting coordinates from Wikipedia articles so they can be displayed in an overlay to making sure oauth works on OSM Wikipedia so we can send editors between the two projects. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On 23 Jul 2009, at 09:17, Ævar Arnfjörð Bjarmason wrote: On Wed, Jul 22, 2009 at 11:34 PM, 80n80n...@gmail.com wrote: ti...@home uses this technique of a basemap with no captions and a text layer for lowzoom tiles (z1 through z6) . Currently they are combined on the server to create a composite image using some GD library I think. This seems to works very well and avoids duplicating the most compute intensive part of generating a map tile. Yeah it would be neat to render a captionless layer in mapnik and then overlay text on it. Can mapnik support that? I'd have thought adding any sort of text to the map might have implications for the rendering itself. The problem is being able to do the text collision avoidance stuff against non-text features, hence why it's not been done before. Shaun smime.p7s Description: S/MIME cryptographic signature ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Shaun McDonald wrote: Yeah it would be neat to render a captionless layer in mapnik and then overlay text on it. Can mapnik support that? I'd have thought adding any sort of text to the map might have implications for the rendering itself. The problem is being able to do the text collision avoidance stuff against non-text features, hence why it's not been done before. Collision avoidance across layers is slightly doable, but requires a good amount of effort. This particular example isn't helped by the fact that you're dealing with variable length texts for the same feature here. -- Lennard ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Thursday 23 July 2009 11:09:32 Shaun McDonald wrote: On 23 Jul 2009, at 09:17, Ævar Arnfjörð Bjarmason wrote: Yeah it would be neat to render a captionless layer in mapnik and then overlay text on it. Can mapnik support that? I'd have thought adding any sort of text to the map might have implications for the rendering itself. The problem is being able to do the text collision avoidance stuff against non-text features, hence why it's not been done before. TopOSM is rendering multiple layers (most of them with Mapnik) and combines them with ImageMagick. http://wiki.openstreetmap.org/wiki/TopOSM -- m.v.g., Cartinus ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 1:25 AM, Aude (Kate)maps2w...@gmail.com wrote: I recommend using OSM data for Brussels, Belgium to test multilingual rendering. The street signs there have street names in both French and Dutch. Likewise, the OSM data contains name:fr and name:nl tags for street names. Thanks everyone, here's a completely unoptimized proof of concept of rendering Brussels in nl/fr: http://cassini.toolserver.org/browse-fr.html http://cassini.toolserver.org/browse-nl.html I'll put up the static renderer the rest of the Wikipedia languages later. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Ævar Arnfjörð Bjarmason avarab at gmail.com writes: http://cassini.toolserver.org/browse-fr.html http://cassini.toolserver.org/browse-nl.html This is very cool. I note that a lot of the tiles are the same, so the disk space requirement wouldn't be too impossible. Do you plan to support 'my main language is English, but I also speak French and Spanish' and so on? There's a huge number of such combinations, but perhaps only a few are common (it would be good to get statistics of Accept-Language http headers on the main osm.org site). -- Ed Avis e...@waniasset.com ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 4:37 PM, Ed Avise...@waniasset.com wrote: Ævar Arnfjörð Bjarmason avarab at gmail.com writes: http://cassini.toolserver.org/browse-fr.html http://cassini.toolserver.org/browse-nl.html The rest are now up at http://cassini.toolserver.org/tile-browse/ This is very cool. I note that a lot of the tiles are the same, so the disk space requirement wouldn't be too impossible. Yeah, hopefully we can cache it with Content-MD5 + Squid. Do you plan to support 'my main language is English, but I also speak French and Spanish' and so on? There's a huge number of such combinations, but perhaps only a few are common (it would be good to get statistics of Accept-Language http headers on the main osm.org site). I'm targeting embedding in Wikipedia articles so I'm just supporting maps in the content language of the wiki itself. That's consistent with every other sort of image / diagram embedded in articles so it's fine for our purposes. Something different could be implemented via user preferences but that would require a new caching key for each new map type, not something I want to target initially at least. ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
From: Ævar Arnfjörð Bjarmason ava...@gmail.com The rest are now up at http://cassini.toolserver.org/tile-browse/ These are great!! (though I think we may just have crashed your renderd with loads of requests) I realize this is just a proof of concept, but is this generally your plan right now, to generate tile sets for every language, with caching? Since the underlying geometries are the same for all tiles, and only the text changes, one thought has been to decouple these into different tile sets (geoms and various localised text tiles), which are then combined on the server before pushing out, or on the client in OL. This would reduce space requirements, and load on the database. Not sure how much more cost it is to overlay the text on geom tiles, on the server, but there are possibly clever ways to make this efficient. There's a need for this on osm.org itself. If an efficient way to localise tiles can be found through your work on wikipedia, it's all very good! -Mikel ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 7:08 PM, Mikel Maronmikel_ma...@yahoo.com wrote: The rest are now up at http://cassini.toolserver.org/tile-browse/ These are great!! (though I think we may just have crashed your renderd with loads of requests) That was part of the plan:) Let's see how it does with added caching though. I realize this is just a proof of concept, but is this generally your plan right now, to generate tile sets for every language, with caching? Yes but on-demand as they're added to pages on Wikimedia projects with the Slippy Map extension. We're not aiming for having a general purpose google-maps-alike in 279 languages but rather just a way for users to embed maps into articles in their language. Those articles will each have their own peephole view of the planet so hopefully we won't have to generate a huge amount of static maps / tiles for each language. Since the underlying geometries are the same for all tiles, and only the text changes, one thought has been to decouple these into different tile sets (geoms and various localised text tiles), which are then combined on the server before pushing out, or on the client in OL. Yeah a basemap + text would be neat. But I couldn't find a way to do it so I thought I'd try the brute-force way first and see how it works. This would reduce space requirements, and load on the database. Not sure how much more cost it is to overlay the text on geom tiles, on the server, but there are possibly clever ways to make this efficient. There's a need for this on osm.org itself. If an efficient way to localise tiles can be found through your work on wikipedia, it's all very good! ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Wed, Jul 22, 2009 at 8:48 PM, Ævar Arnfjörð Bjarmason ava...@gmail.comwrote: On Wed, Jul 22, 2009 at 7:08 PM, Mikel Maronmikel_ma...@yahoo.com wrote: The rest are now up at http://cassini.toolserver.org/tile-browse/ These are great!! (though I think we may just have crashed your renderd with loads of requests) That was part of the plan:) Let's see how it does with added caching though. I realize this is just a proof of concept, but is this generally your plan right now, to generate tile sets for every language, with caching? Yes but on-demand as they're added to pages on Wikimedia projects with the Slippy Map extension. We're not aiming for having a general purpose google-maps-alike in 279 languages but rather just a way for users to embed maps into articles in their language. Those articles will each have their own peephole view of the planet so hopefully we won't have to generate a huge amount of static maps / tiles for each language. Since the underlying geometries are the same for all tiles, and only the text changes, one thought has been to decouple these into different tile sets (geoms and various localised text tiles), which are then combined on the server before pushing out, or on the client in OL. Yeah a basemap + text would be neat. But I couldn't find a way to do it so I thought I'd try the brute-force way first and see how it works. ti...@home uses this technique of a basemap with no captions and a text layer for lowzoom tiles (z1 through z6) . Currently they are combined on the server to create a composite image using some GD library I think. This seems to works very well and avoids duplicating the most compute intensive part of generating a map tile. Etienne This would reduce space requirements, and load on the database. Not sure how much more cost it is to overlay the text on geom tiles, on the server, but there are possibly clever ways to make this efficient. There's a need for this on osm.org itself. If an efficient way to localise tiles can be found through your work on wikipedia, it's all very good! ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Looks very nice! Are there any further plans to translate the maps based on wikipedia data? Would that various names be better imported into main OSM db or just into the DB for rendering? Will this translation be done based only on language links in existing articles or also based on lists of exonyms [1], which would cover also places without wikipedia articles? Can we help in some way? Stefan [1] http://en.wikipedia.org/wiki/Category:Exonyms On Thu, Jul 23, 2009 at 1:34 AM, 80n80n...@gmail.com wrote: On Wed, Jul 22, 2009 at 8:48 PM, Ævar Arnfjörð Bjarmason ava...@gmail.com wrote: On Wed, Jul 22, 2009 at 7:08 PM, Mikel Maronmikel_ma...@yahoo.com wrote: The rest are now up at http://cassini.toolserver.org/tile-browse/ These are great!! (though I think we may just have crashed your renderd with loads of requests) That was part of the plan:) Let's see how it does with added caching though. I realize this is just a proof of concept, but is this generally your plan right now, to generate tile sets for every language, with caching? Yes but on-demand as they're added to pages on Wikimedia projects with the Slippy Map extension. We're not aiming for having a general purpose google-maps-alike in 279 languages but rather just a way for users to embed maps into articles in their language. Those articles will each have their own peephole view of the planet so hopefully we won't have to generate a huge amount of static maps / tiles for each language. Since the underlying geometries are the same for all tiles, and only the text changes, one thought has been to decouple these into different tile sets (geoms and various localised text tiles), which are then combined on the server before pushing out, or on the client in OL. Yeah a basemap + text would be neat. But I couldn't find a way to do it so I thought I'd try the brute-force way first and see how it works. ti...@home uses this technique of a basemap with no captions and a text layer for lowzoom tiles (z1 through z6) . Currently they are combined on the server to create a composite image using some GD library I think. This seems to works very well and avoids duplicating the most compute intensive part of generating a map tile. Etienne This would reduce space requirements, and load on the database. Not sure how much more cost it is to overlay the text on geom tiles, on the server, but there are possibly clever ways to make this efficient. There's a need for this on osm.org itself. If an efficient way to localise tiles can be found through your work on wikipedia, it's all very good! ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
On Tue, 21 Jul 2009 22:05:40 +, Ævar Arnfjörð Bjarmason ava...@gmail.com wrote: I'm doing some experiments with multilingular rendering (for deploying on Wikimedia sites, see Maps-l) but I'm lacking good test data. What areas on the OSM map are especially rich in name:$code tags for different languages? Preferably down to street level. I would suggest any multi-lingual conflict zones such as Cypros (greek, turkish, english), and countries with several official languages (Belgium). Should be some good opertunities in south-east asia, but do not know how good the data is. Cyprus/Belgium should have a good dataset to work with. I thought Gaza might be pretty good but it mainly just has an odd mix of name:en and name= where the main name tag contains the arabic version too. There should be a name:en or similar for every name written with non-latin characters Are there any areas that are better? Preferably with more than two languages. Try Quibec (Frensh/English), Cypros (Greek/Turkish/English?), Belgium (French/German/Dutch), Switzerland (French/German/Italian/?), Southern Finland (Finnish/Swedish), Northern Norway/Northern Sweden/Northern Finland (Norwegian/Swedish/Finnish/various Samii names), Greenland might also be a source (Danish/Inuit), check also out Thailand, Malaisia, Singapore, Bangladesh, Sri-Lanka, Hong Kong where the names might be written with various types of lettering (Latin/Sanskrete/Chinese) even if the names are the same in all languages. Besides, important cities, such as national capitals should also be available in all variations of the name (Peking/Pequim/Bei Jing/++ for the chinese capital). ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk -- Brgds Aun Johnsen via Webmail ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] i18n-rich areas on the map
Aude (Kate) schreef: I recommend using OSM data for Brussels, Belgium to test multilingual rendering. The street signs there have street names in both French and Dutch. Likewise, the OSM data contains name:fr and name:nl tags for street names. I've been in Southern France and saw some villages with street names in French, but also some south french dialect, close to spanish. Could inspire people, it was in Montsalvy. south of Aurillac. In Aurillac, there is an office, to defend that dialect... Marc -- Shortwave transmissions in English, Francais, Nederlands, Deutsch, Suid-Afrikaans, Chinese, Dansk, Urdu, Cantonese, Greek, Spanish, Portuguese, ... http://users.fulladsl.be/spb13810/swlist/ http://radiolanguages.tk Stations list: http://users.fulladsl.be/spb13810/txlist/ ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
