Translation tools in Booktype
  • Anybody needs translation tools in Booktype?
  • 33 Comments sorted by
  • Vote Up0Vote Down adamadam
    Posts: 88Member
    there is a selftranslate python module

    https://github.com/tuukka/selftranslate

    that was developed for booktype and plug straight in but i think it hasnt
    been tested with the newest booktype and also the interface could do with
    improvement


    adam

    On Tue, Jun 12, 2012 at 7:35 AM, Boštjan Jerko <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > Anybody needs translation tools in Booktype?
    >
    >
  • If I understand correctly selftranslate is a tool to translate the environment.
    I am talking about book translation tools (e.g. SDL trados or Wordfast). So a tool to help you translate a book - with term memory ...
  • Vote Up0Vote Down adamadam
    Posts: 88Member
    ah yes

    we do need to integrate a better translation process. I think we need two
    types of process:
    1. an easy 'community' process for translation
    2. a workflow that could integrate with other tools for professional
    translators

    the first should be built in, and the second would be an integration import
    / export flow I am guessing.

    While both are interesting I think the first one is most interesting since
    there is a lot of need for community translation with a very light workflow.

    What were you thinking?


    adam

    On Tue, Jun 12, 2012 at 2:39 PM, Boštjan Jerko <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > If I understand correctly selftranslate is a tool to translate the
    > environment.
    > I am talking about book translation tools (e.g. SDL tradosor
    > Wordfast ). So a tool to help you translate a
    > book - with term memory ...
    >
    >
  • Also on the topic of translation of content, I'd like to repeat one of the
    main requests coming from One Laptop Per Child: Making it easier to handle
    screenshots in the translations. Booktype's current image handling makes it
    very difficult to work with screenshots (and I suppose all images) in
    translated versions.

    doug



    On Tue, Jun 12, 2012 at 2:44 PM, Adam Hyde <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > ah yes
    >
    > we do need to integrate a better translation process. I think we need two
    > types of process:
    > 1. an easy 'community' process for translation
    > 2. a workflow that could integrate with other tools for professional
    > translators
    >
    > the first should be built in, and the second would be an integration import
    > / export flow I am guessing.
    >
    > While both are interesting I think the first one is most interesting since
    > there is a lot of need for community translation with a very light
    > workflow.
    >
    > What were you thinking?
    >
    >
    > adam
    >
    >
    > On Tue, Jun 12, 2012 at 2:39 PM, Boštjan Jerko <<br />> booktype-dev@lists.sourcefabric.org> wrote:
    >
    > > If I understand correctly selftranslate is a tool to translate the
    > > environment.
    > > I am talking about book translation tools (e.g. SDL tradosor
    > > Wordfast ). So a tool to help you translate a
    >
    > > book - with term memory ...
    > >
    > >
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • Vote Up0Vote Down adamadam
    Posts: 88Member
    what do you think of drag and drop uploading of images? ie. drag an image
    to the page and it is placed and uploaded at the same time

    adam

    On Tue, Jun 12, 2012 at 2:49 PM, Douglas Arellanes <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > Also on the topic of translation of content, I'd like to repeat one of the
    > main requests coming from One Laptop Per Child: Making it easier to handle
    > screenshots in the translations. Booktype's current image handling makes it
    > very difficult to work with screenshots (and I suppose all images) in
    > translated versions.
    >
    > doug
    >
    >
    >
    >
    > On Tue, Jun 12, 2012 at 2:44 PM, Adam Hyde <<br />> booktype-dev@lists.sourcefabric.org> wrote:
    >
    > > ah yes
    > >
    > > we do need to integrate a better translation process. I think we need two
    > > types of process:
    > > 1. an easy 'community' process for translation
    > > 2. a workflow that could integrate with other tools for professional
    > > translators
    > >
    > > the first should be built in, and the second would be an integration
    > import
    > > / export flow I am guessing.
    > >
    > > While both are interesting I think the first one is most interesting
    > since
    > > there is a lot of need for community translation with a very light
    > > workflow.
    > >
    > > What were you thinking?
    > >
    > >
    > > adam
    > >
    > >
    > > On Tue, Jun 12, 2012 at 2:39 PM, Boštjan Jerko <<br />> > booktype-dev@lists.sourcefabric.org> wrote:
    > >
    > > > If I understand correctly selftranslate is a tool to translate the
    > > > environment.
    > > > I am talking about book translation tools (e.g. SDL tradosor
    > > > Wordfast ). So a tool to help you translate a
    > >
    > > > book - with term memory ...
    > > >
    > > >
    > >
    > >
    >
    >
  • Drag and drop of images would help, certainly.

    The translation workflow should have a more elegant workflow for image
    handling. It would be more useful to have some kind of mechanism where
    Booktype would know:

    - which images are used in the original version
    - the user would be prompted to substitute new images for the old ones
    - on upload, the user would be prompted to keep the positioning of the
    images as in the original

    In addition, batch uploading should be enabled; in OLPC's case, all their
    screenshots are automatically generated based on a single script, which
    incidentally creates all the different language versions of their UI
    elements automatically. If they were able to simply batch upload all their
    screenshots, then they could go through such a prompt process and replace
    the original images with the translated ones in a single go.


    doug





    On Tue, Jun 12, 2012 at 2:50 PM, Adam Hyde <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > what do you think of drag and drop uploading of images? ie. drag an image
    > to the page and it is placed and uploaded at the same time
    >
    > adam
    >
    >
    > On Tue, Jun 12, 2012 at 2:49 PM, Douglas Arellanes <<br />> booktype-dev@lists.sourcefabric.org> wrote:
    >
    > > Also on the topic of translation of content, I'd like to repeat one of
    > the
    > > main requests coming from One Laptop Per Child: Making it easier to
    > handle
    > > screenshots in the translations. Booktype's current image handling makes
    > it
    > > very difficult to work with screenshots (and I suppose all images) in
    > > translated versions.
    > >
    > > doug
    > >
    > >
    > >
    > >
    > > On Tue, Jun 12, 2012 at 2:44 PM, Adam Hyde <<br />> > booktype-dev@lists.sourcefabric.org> wrote:
    > >
    > > > ah yes
    > > >
    > > > we do need to integrate a better translation process. I think we need
    > two
    > > > types of process:
    > > > 1. an easy 'community' process for translation
    > > > 2. a workflow that could integrate with other tools for professional
    > > > translators
    > > >
    > > > the first should be built in, and the second would be an integration
    > > import
    > > > / export flow I am guessing.
    > > >
    > > > While both are interesting I think the first one is most interesting
    > > since
    > > > there is a lot of need for community translation with a very light
    > > > workflow.
    > > >
    > > > What were you thinking?
    > > >
    > > >
    > > > adam
    > > >
    > > >
    > > > On Tue, Jun 12, 2012 at 2:39 PM, Boštjan Jerko <<br />> > > booktype-dev@lists.sourcefabric.org> wrote:
    > > >
    > > > > If I understand correctly selftranslate is a tool to translate the
    > > > > environment.
    > > > > I am talking about book translation tools (e.g. SDL tradosor
    > > > > Wordfast ). So a tool to help you translate a
    > > >
    > > > > book - with term memory ...
    > > > >
    > > > >
    > > >
    > > >
    > >
    > >
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • @Adam: agree with "community" translation tools. My wife did a short stint as a translator and tried to use one of the more powerful tools. Had a lot of troubles for a "one woman band" translation agency. So I am definitely in the let's start small camp.

    @Doug: Hi :).. I wasn't thinking about images, but yes you are right. There should be some sort of flag for the images in the original and translation. I guess there should be some sort of a top "flag" to know which is original and then tree structure of translations with appropriate images.
  • Regarding Booktype localization: 

    I am not exactly sure about this, I have ended up doing translation straight into the templates in both Booki and Booktype, since the site is pretty much only for one language, and there will be small changes to the templates in any case. 

    But the issue might be a lot more complicated for a multi-language Booktype.

    About translation tools for books:

    I don't really know, never actually bothered to use anything like this. 

    It would make sense to take the localization file of a program and then use that as a translation memory, so basically the things that are in the interface would already be translated when you start. 

    Isn't this more of a question about TinyMCE? I haven't even checked yet if it would be possible to use a Finnish proof reading module with it. 
  • Tomi
    I'm not sure if I understand you correctly.
    I was talking about tool where you have some sort of database of already done translations (mostly phrases etc.) so you don't bang your head with repeated phrases. Not to mention change the translation in the course of it.
    So you can keep consistent terminology and also (for "let's call it professional tools") share translated terms with other users.
  • Bostjan, as a longtime translator and Wordfast user, I know exactly what
    you are talking about.

    Tomi (and others), a translation memory is a tool that checks individual
    sentences in the text of an article (or book) and compares them to
    sentences the translator has previously entered. This is extremely helpful
    in many situations, especially when dealing with laws and other repetitive
    texts where the same phrases are used many times.

    As a book translator (the German translation of Tomas Sedlacek's "Economics
    of Good and Evil," based on my English translation, is still on the
    bestseller lists in Germany and Switzerland,
    http://www.perlentaucher.de/buch/tomas-sedlacek/die-oekonomie-von-gut-und-boese.html),
    Wordfast has meant my translating times are cut by more than half.

    I'd suggest taking a look at www.wordfast.net for an example of how
    Wordfast works - I believe they have a trial version - and it runs fine on
    Linux.


    doug



    On Tue, Jun 12, 2012 at 5:42 PM, Boštjan Jerko <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > Tomi
    > I'm not sure if I understand you correctly.
    > I was talking about tool where you have some sort of database of already
    > done translations (mostly phrases etc.) so you don't bang your head with
    > repeated phrases. Not to mention change the translation in the course of it.
    > So you can keep consistent terminology and also (for "let's call it
    > professional tools") share translated terms with other users.
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • @doug Not sure if Wordfast works in Linux since it looks like it's tied to Word. It works on Mac though, so I'm checking it ... again :D
  • Oops. I'm wrong. There is a Linux version.
  • Yep, the whole thing is based on Eclipse, and it works just the same across
    platforms. I've been using it on my Ubuntu install for a couple of years
    now.

    doug


    On Tue, Jun 12, 2012 at 9:04 PM, Boštjan Jerko <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > Oops. I'm wrong. There is a Linux version.
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • What I did for a book I needed translated from English to Spanish that I was editing in LyX (as that was borderline easy enough for the editors to handle) was to let LyX give me latex output of the English version. Then I studied how Google Translate would react to different parts of latex syntax and would try to make these "survive" a translation by substituting before and after a translation run. The resulting Latex file I would import into Lyx again. I still needed to go through the result in order to clean up technical glitches and in order to make the Spanish understandable enough for the Spanish editors to be able to work on it. It still saved me a lot of work in comparison with me having translated the entire text by myself.

    Maybe something similar could be done in the form of different plugins for the various translation services? The code would likely have to be tweaked constantly, as the existing translation services likely change how they handle different symbols...
  • One of the many problems with the above script is that Google Translate doesn't recognize that something is just one sentence when it is interrupted by some controlling keyword. and then, Google Translate is not free software, of course...
  • I see Google Translate as a part of the tool. You can use it as a helping tool, but not for a translation.
    For example I am currently translating a book from English to Slovene and need to translate: Achilles tendonitis or Shin splints and it would be nice if I could use the same term for the phrases.

    That's when translation memory (as Doug said) would come handy.
  • What do you mean by "You can use it as a helping tool, but not for a translation."? What else would you use it as a helping tool for?

    Different use cases may make different translation tools more useful. For laws, and instructions, it may make sense to build up your own translation memory. "click here" and "move on to the next page" and similar phrases will likely be commonly used in for example computer documentation.

    For literature and texts in humanities and social sciences, which focus on not sounding repetitive, this may not really be possible and you would want a translation memory more of the size that Google Translate and other services with a massive user base provide (is there a free software based service I am unaware of?).
  • I can tell you that for Slovene Google translate is still rubbish. I never use it to translate the whole text, but to check translation of separate words.
    My dad doesn't understand English so he wanted to use GT, but resorted to asking me to translate the text since it was of very poor quality.
    Mind you I'm not bashing GT as a tool - it's great, but we'll have to wait for some time to be really useful as a language translation tool - but I guess it depends on the languages (and size of the corpora).
  • Reply to @Douglas+Arellanes:

    Making it easier to handle
    screenshots in the translations.



    in case this sparks some ideas, quote from the SVG Embed module for Drupal:

    What this module does is to provide a text filter that
    automatically embeds an SVG graphic into your html text and before doing
    this, translates all the textstrings in the graphic. This has huge
    advantages on multi lingual sites: the graphic always adjusts to the
    surrounding content or the user preferences and gets displayed in the
    correct language.

    I suppose that for screenshots, first we would require a tool for taking SVG screenshots that preserve the text as text. Or maybe there's another way. I don't know enough about display managers to know if that's possible or not. Getting more complicated: in the SVG screenshots, put the text string identifier, and then in Booktype, or some other intermediary, pull in the most up-to-date version of that string in the correct language... hmmm, perhaps a problem if the length of the string changes significantly, thus changing the display layout (such as the size of a button) between languages and versions.

    Post edited by Patrick Gibbs at 2012-09-10 23:18:34
  • Reply to @Douglas+Arellanes:

    Making it easier to handle
    screenshots in the translations.



    in case this sparks some ideas, quote from the SVG Embed module for Drupal:

    What this module does is to provide a text filter that
    automatically embeds an SVG graphic into your html text and before doing
    this, translates all the textstrings in the graphic. This has huge
    advantages on multi lingual sites: the graphic always adjusts to the
    surounding content or the user preferences and gets displayed in the
    correct language.

    I suppose that for screenshots, first
    we would require a tool for taking SVG screenshots that preserve the
    text as text. Or maybe there's another way. I don't know enough about
    display managers to know if that's possible or not.


  • Vote Up0Vote Down Sava TatićSava Tatić
    Posts: 113Member, Administrator, Sourcefabric Team
    Patrick, thanks for pointing out the module. Sounds very good.

    Johannes, re: translation memory, no matter what you translate, be it humanities or technical/legal texts, you want consistency. As a translator, you want to know what term you have been using over the course of e.g. novel for a particular type of say shovel they use in let's say the Brazilian Northeast. It also makes it easier for a translator to go back and change if she's actually found a better solution (translation) for a particular phrase or word. So by all means, we need a translation memory, regardless of whether you can get suggestions from Google Translate or any other automaton. 

    In any case, either machine translation or translation memory should be just aids to a living, breathing translator(s), presenting suggestions that one would accept or reject. It's great to have tools like GT evolving, and we should try to build plugs for any number of them as time goes, but we must implement a translation memory if we want to be serious about quality translation. And users translation memories could either be aggregated on a server (if one chooses to share) or  peer-to-peer. 

    I haven't done any research recently about what's out there (I remember there was an Open Office plugin), but research we should. If we don't find a way, we'll make one. 
    Post edited by Sava Tatić at 2012-09-11 03:55:30
  • The open source translation memory I've had my eye on for some time now is
    the Worldwide Lexicon, but on the commercial side Wordfast has been doing
    interesting things with an online TM too.


    Doug


    On Tuesday, September 11, 2012, Sava Tatić wrote:

    > Patrick, thanks for pointing out the module. Sounds very good.
    >
    > Johannes, re: translation memory, no matter what you translate, be it
    > humanities or technical/legal texts, you want consistency. As a translator,
    > you want to know what term you have been using over the course of e.g.
    > novel for a particular type of say shovel they use in let's say the
    > Brazilian Northeast. It also makes it easier for a translator to go back
    > and change if she's actually found a better solution (translation) for a
    > particular phrase or word. So by all means, we need a translation memory,
    > regardless if you can get suggestions from Google Translate or any other
    > automaton.
    >
    > In any case, either machine translation or translation memory should be
    > just aids to a living, breathing translator(s), presenting suggestions that
    > one would accept or reject. It's great to have tools like GT evolving, and
    > we should try to build plugs for any number of them as time goes, but we
    > must implement a translation memory if we want to be serious about quality
    > translation. And users translation memories could either be aggregated on a
    > server (if one chooses to share) or peer-to-peer.
    >
    > I haven't done any research recently about what's out there (I remember
    > there was an Open Office plugin), but research we should. If we don't find
    > a way, we'll make one.
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • Vote Up0Vote Down Sava TatićSava Tatić
    Posts: 113Member, Administrator, Sourcefabric Team
    Worldwide Lexicon looks interesting. Has anyone taken a closer look?
  • On Tue, Sep 11, 2012 at 12:54 AM, Sava Tatić <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > Patrick, thanks for pointing out the module. Sounds very good.
    >
    > Johannes, re: translation memory, no matter what you translate, be it
    > humanities or technical/legal texts, you want consistency. As a translator,
    > you want to know what term you have been using over the course of e.g.
    > novel for a particular type of say shovel they use in let's say the
    > Brazilian Northeast. It also makes it easier for a translator to go back
    > and change if she's actually found a better solution (translation) for a
    > particular phrase or word. So by all means, we need a translation memory,
    > regardless if you can get suggestions from Google Translate or any other
    > automaton.
    >
    > In any case, either machine translation or translation memory should be
    > just aids to a living, breathing translator(s), presenting suggestions that
    > one would accept or reject. It's great to have tools like GT evolving, and
    > we should try to build plugs for any number of them as time goes, but we
    > must implement a translation memory if we want to be serious about quality
    > translation. And users translation memories could either be aggregated on a
    > server (if one chooses to share) or peer-to-peer.
    >
    >
    I forgot what the discussion I opinionated in was about, and I cannot
    refind my own post. :)

    My point may have been that we won't be able to compete with Google
    Translate, especially if we do this as a side product hidden in some menu
    of Booktype. It would be hard to get the same level of user contributions
    which enhances the service as Google Translate and similar services do.
    Also Google Translate has some memory of previous translation it gave you,
    and I would assume that they work with per-user translation memories.
    Having an open solution would be of long-term benefit to the world, it
    would seem to me, but I would imagine that it would have to be independent
    of any other software in order to be successful.


    > I haven't done any research recently about what's out there (I remember
    > there was an Open Office plugin), but research we should. If we don't find
    > a way, we'll make one.
    >
    >
  • Hi,


    To continue the discussion, I noticed on a different mailing list that
    Brian McConnell, one of the founders of the Worldwide Lexicon, has a new
    project called Xlatn, which is a series of resources for translators,
    localizers and software developers. I found their developer page to be
    quite useful in bringing together a lot of different material.

    http://xlatn.com/for-developers/

    Hope this helps!


    doug



    On Tue, Sep 11, 2012 at 12:57 PM, Johannes Wilm <<br />booktype-dev@lists.sourcefabric.org> wrote:

    > On Tue, Sep 11, 2012 at 12:54 AM, Sava Tatić <<br />> booktype-dev@lists.sourcefabric.org> wrote:
    >
    > > Patrick, thanks for pointing out the module. Sounds very good.
    > >
    > > Johannes, re: translation memory, no matter what you translate, be it
    > > humanities or technical/legal texts, you want consistency. As a
    > translator,
    > > you want to know what term you have been using over the course of e.g.
    > > novel for a particular type of say shovel they use in let's say the
    > > Brazilian Northeast. It also makes it easier for a translator to go back
    > > and change if she's actually found a better solution (translation) for a
    > > particular phrase or word. So by all means, we need a translation memory,
    > > regardless if you can get suggestions from Google Translate or any other
    > > automaton.
    > >
    > > In any case, either machine translation or translation memory should be
    > > just aids to a living, breathing translator(s), presenting suggestions
    > that
    > > one would accept or reject. It's great to have tools like GT evolving,
    > and
    > > we should try to build plugs for any number of them as time goes, but we
    > > must implement a translation memory if we want to be serious about
    > quality
    > > translation. And users translation memories could either be aggregated
    > on a
    > > server (if one chooses to share) or peer-to-peer.
    > >
    > >
    > I forgot what the discussion I opinionated in was about, and I cannot
    > refind my own post. :)
    >
    > My point may have been that we won't be able to compete with Google
    > Translate, especially if we do this as a side product hidden in some menu
    > of Booktype. It would be hard to get the same level of user contributions
    > which enhances the service as Google Translate and similar services do.
    > Also Google Translate has some memory of previous translation it gave you,
    > and I would assume that they work with per-user translation memories.
    > Having an open solution would be of long-term benefit to the world, it
    > would seem to me, but I would imagine that it would have to be independent
    > of any other software in order to be successful.
    >
    >
    >
    > > I haven't done any research recently about what's out there (I remember
    > > there was an Open Office plugin), but research we should. If we don't
    > find
    > > a way, we'll make one.
    > >
    > >
    >
    >
    Douglas Arellanes
    Director of Innovation
    Sourcefabric, o.p.s.

    Find a way or make one.
  • Vote Up0Vote Down Sava TatićSava Tatić
    Posts: 113Member, Administrator, Sourcefabric Team
    Johannes,

    I wasn't talking about competing with Google Translate. I am talking about something almost completely different. 

    For a professional translator, the most important thing is the record (memory) of solutions (terms, phrases, etc.) he has used before. For a group of translators, and Booktype is foremost about collaboration, their collective memory as a group is the most important thing. It ensures consistency, saves time, and provides iterative improvements to the overall translation quality. The reason is you trust no one as much as you do trust yourself and the peers who are working with you on a particular project. You don't depend on any whim of the algorithm or translation memory overload from questionable sources. You know what you have there is correct and the way you prefer it, and if you have better ideas, you have a way of correcting it. 

    Providing such a functionality is not rocket science, and there is code out there as we have seen. It will never be a Google Translate killer, nor is it intended to be. It's a way of helping people help themselves as they build their own body of translated work. 

    (BTW, you can always find the thread on the forum, see the link below this post :)



    Post edited by Sava Tatić at 2012-09-11 17:03:44
  • yes, I agree when it comes to application translation. I was the translator
    of Biblatex, Scribus and Shotwell and a number of other programs I don't
    remember to Danish/Norwegian in the past and having a tool that would let
    me know what word I used for "file" and "close", etc. was very helpful. I
    agree that exists and is nothing new.

    When it comes to longer texts, such as books, it is a little different. I
    have to translate not only individual terms, but sentence structure and
    longer phrases. I have to make sure that my language is varied, that I
    don't use the same adjective two or three times in a paragraph.

    That's why Google Translate comes in handy. Even though the output is
    garbled, it is still easier to "fix" this language than to do it all by
    hand. My friends who are working only in translation therefore translate
    first with such an online tool, and then adjust the language. But maybe
    this is different depending on the language pair and how good the Google
    translation is?


    I noticed this project: http://tatoeba.org/eng/ It seems to be in it's
    infancy, but it seems to be based upon sentence translation and if it's not
    in there already, it should be extendable with a personal translation
    memory.

    Additionally one has to make adjustment for cultural differences and
    knowledge of target audience. An extreme example may be: 

    Danish:
    Waking up way late at my girlfriend's house in Nørrebrograde, I dashed out
    into the sun and across Strøget, passing the reds and whites of this 5th of
    June and arrived just 5 minutes late at Super Brugsen. 

    English:
    Although we had been together for five years, Danish custom meant that
    marriage was not an issue, and it was not strange for me to sleep at my
    girlfriend's house situated in central Copenhagen in a district that had
    previously been known to house mainly drunks, but now was up and coming. I
    woke up quite late that morning and so I had to sprint across the shopping
    street where people were celebrating constitution day on this 5th of June.
    I managed to get to the store just 5 minutes late. 
    Post edited by Johannes Wilm at 2012-09-11 17:59:57
  • @Sava: Ah yes, I had forgotten I had posted in this forum. I was looking through my mail tying to find it instead.

    http://www.worldwidelexicon.org/ looks interesting, but the link to the blog looks like it's dead. but as far as what I can gather, it is a service where translators can provide translations for requests that come in. And those translators will then in turn use Google Translate to get a rough version, I would assume. As I understand the way their personal translation memory works is that if the same sentence/paragraph is requested twice by the same user, it will give the exact same result. The trick is that one an then have a website that has English as it's standard language, but offers 20 other languages. If anyone clicks on another language of a specific article, the user is presented with a machine translation of the page (via Google Translate or similar). If over for example 20 people request that article in one specific language, a request is made to have it translated by a human being at 6 cents/word. If another user clicks on the link after that, this "translation memory" kicks in and the user gets presented with the translation tat was already made once.

    For that application it works well. The other type of "translation memory" seems to be based on individual words, right? That's the type I can use when translating the interface of an application. I also agree that it could come in handy when encountering specific technical terminology that needs to be the same throughout a book. But I would really need something more to give me a rough translation of the text. Tatoeba may be it.
  • http://xlatn.com/for-developers/  makes the distinction between the two types of translation quite explicit and I don't see any automatic tools for the translation of dynamic content there. They seem to recommend the same as I do: use a web service and then improve it with human beings.
  • Vote Up1Vote Down Sava TatićSava Tatić
    Posts: 113Member, Administrator, Sourcefabric Team
    Fixing up a bad translation or doing it from scratch, that's the eternal debate among translators. It's akin to who is stronger, Bruce Lee or Chuck Norris. 

    For me it always depended on how bad a translation is. If we are talking high-brow literature or humor, a rough translation by an automaton is practically useless. For weather, stocks and the like, machines can do a good job. 

    For the languages I know well, I would never use Google Translate to translate a whole chapter then go fix it up. And I use it every day as a quick reference tool. For the languages I don't know well, I would use Google Translate to cross-check with my own comprehension (but I'd never do any real translation work in such languages, so it's a moot point) :)


  • Yeah, it may come down to personal preferences and the languages involved. :) 
  • Worldwide Lexicon looks like it's a useful tool, but it doesn't seem to work for me. I've tried to use it to translate a simple word, but nothing (with using machine translation, of course).
    I'll still keep an eye on it...
  • Reply to @Patrick+Gibbs:

    I also see this about the Transposh plugin for Wordpress:

    Is it possible to use different images in different languages in my themes?


    Yes, although a bit tricky – you can either use the $my_transposh_plugin-&gt;target_language as
    part of the image descriptor, this will load different images based on
    the current language. Alternatively, in places where shortcodes are
    supported you can use the mylang shortcode as in the following example:

    <br>
    &lt;img src="http://s.transposh.net/s/logo[tp mylang="y"
    lang="he,es"][/tp].png" title="transposh logo" width="300" height="86"
    /&gt;<br><br>above from http://transposh.org/faq/<br&gt;