All, As part of the MosesCore project we, at Capita TI formerly ALS, have designed a proposed open standard API for MT systems. This specifies a service interface rather than an application interface where the intention was to make MT systems of different flavours look the same. Applications, then, can use *any* compliant MT back-end since they have a consistent interface. This approach allows the application to define all of the application concerns, e.g. security, transport, application framework, whilst defining a highly cohesive and low coupled framework for the back-end MT systems.
The standard is called OMTC: Open Machine Translation Core. There is a reference implementation available on GitHub at git://github.com/ianj-als/omtc.git Please clone, or fork this repo to your heart's content. It would be good to get your thoughts on this and how it could be made better. Any comments or constructive feedback can be sent straight to me at [email protected] Thanks, Ian On Tue, 12 Feb 2013 08:09:06 -0000, Barry Haddow <[email protected]> wrote: > Hi Tom > > As far as I know, REST > (http://en.wikipedia.org/wiki/Representational_state_transfer) is a type > of software architecture, rather than an API, and I think the XML-RPC > interface could be considered RESTful. > > As you're probably aware, the Moses server alone isn't sufficient to > create a "translation system". In the past we have also used the Moses > server, along with tokenisation and casing tools, to create a > translation service implementing the (http-based) Google translation > API, for use in some experiments. I haven't looked in any detail at the > TAUS API, but if it's to become an industry standard then it may be > useful to have a Moses wrapper implementing the API. This wouldn't > necessarily have to be part of Moses though -- it could be fairly > loosely coupled. > > As to removing the Moses server, I know this was mentioned a while back, > but I can't see it happening unless there is something equivalent but > better to take its place. In the CAT projects that Edinburgh is involved > in (MateCat and Casmacat) there's a need for a server version of Moses, > and instead of using the existing server, they wrote a new server based > in CherryPy which better supports batching of requests. There's some > discussions at the moment about extra functionality required by these > projects, so the outcome of that may be a replacement for Moses server, > but it's too early to say, > > cheers - Barry > > On 11/02/13 17:15, Tom Hoar wrote: >> >> Moses' mosesserver supports XML-RPC protocol for translations. There >> was some mention ~6 months ago that this might not continue to be >> supported. Has there been any further consideration of this? >> >> Also, TAUS (a MosesCore partner) mentions that the REST API has become >> popular for some translation environments. Would it be a viable >> alternative to or replacement for the XML-RPC layer? >> >> >> >> _______________________________________________ >> Moses-support mailing list >> [email protected] >> http://mailman.mit.edu/mailman/listinfo/moses-support > > -- Kind regards, Ian Johnson Software Engineer Capita Translation and Interpreting Riverside Court, Huddersfield Road, Delph, Oldham, OL3 5FZ | Tel (UK): +44 845 367 7000 | Tel (US): +1 (800) 579-5010 | [email protected] | Skype ID: ian.johnson_als www.capitatranslationinterpreting.com _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
