Brian Suda wrote: Wiktionary's database structure is the same as Wikipedia's: A bunch of free-form bloks of text using MediaWiki mark-up syntax. There is no real way to extract any algorithmically useful data, such as "X is a translation of Y into language Z" or anything of the sort. I've been saying that since Wiktionary started. :-)

Brian Suda wrote: Hoi, I read your mail with intrest. It made me look into what the .dict format is. It is described in RFC 2229. It allows for people to look up information from their computer and the information is delivered from one of the many hosts that may hold the requested information. As Angela said in her reply to your post, we are working on a new iteration of Wiktionary that is going by the name of "Ultimate Wiktionary". This will have a relational database at its heart. It is intended to have content in all languages and the first challenge is to make it work in the first place. The second challenge is to create a User Interface that translates to all these languages and the third challenge is to have an import and export mechanism, preferably using a standards based XML scheme. We hope to show something at the Wikimania event. In your mail you want to export the wiktionary data. The consequence is that when you choose for export, you will have to do this continually as we hope to increase the content of Ultimate Wiktionary dramatically. As far as I understand the RFC, there is a need for responding to a request and providing a reply in a set format. There is no need to have a database in a specific format as long as the respons provided conforms to the RFC. As the Ultimate Wiktionary is being designed at the moment and as Wikidata is being built, this is a time to consider what is needed to provide .dict functionality. This functionality will be included when someone does the programming or when someone finds the funds to strap the .dict functionality on top of Ultimate Wiktionary. At this time it is premature to think about exporting from UW as UW has not been built yet. It will however be possible to do so. Some of the current wiktionaries can be parsed into information as they are highly structured. Some will prove an interesting challenge to convert to any other format. Because of its lack of structure and consistency it is as closed as any proprietary format would be. Even the names of an article is not necessarily the name of the associated word as some Wiktionaries still capitalise the first character of a word. One problem I see with exporting content from Wiktionary is the GNU-FDL requirement to maintain the history of the contributors. For the UW, I think it can be solved by adding the history information on the talk page. The necessity for UW stems from the likelyhood that many wiktionaries, if not all, may merge into the Ultimate Wiktionary and be abandoned. Thanks, GerardM.

Brian Suda wrote: In fact exporting the Wiktionaries is almost impossible right now. I once tried to write a script in Python to convert an English Wiktionary entry in some common, logical format. There are too many different possible ways an entry can be built up. Maybe I'll try it again one day, but my time has become very limited lately. I'm not saying it's entirely impossible, but any solution will need to have some manual input. It's very hard to automate it all the way. Another thing is that the Wiktionary content is mostly not ready yet. We've done an amazing amount of work already and some entries are already quite good, but most of the content needs a few (maybe 10 or 20) more years of work to be able to say before it will be usable by the public. We should be thinking about making it possible to export to various formats right now though. If all you want to to is look up a word and keep whatever is currently on its page and simply present that to the user. (or only grab the part describing the English word) That should be possible and would be very easy to do. Maybe I should look into this dict format, to know whether that would work. I'll go do that right now... Jo