Ne ineydamides arnaelidora ys rai nideyriel elnageyigi a' naaser, kenaws ineyn mikani arnaelidora (MA), aasbokielelyr nereyoneeyn sleykkdyreyel mageelang ys ne grimmireyn ys ne elnageyigdy nas ne neelidorasob vedwere ddem. Riser sna slnag-kagang ddee grimmireyn nas neelidorasob, slami mikani arnaelidora e ardeyn embelayr gidi-grovre mesag, weene ne gaiel areyn a' eleirs rram in elirgi inmaeynd ys arianang simbeldy ys inkkeyridi arnaelidora. Rai sleyk gidi-grovre iniraik areyn didodokiel MA, weene elnageyige nas inelognmred mageeleyn inne ineydamidokielelyr angeykes rram birieleleel karbari. E'nish, wark slieyn inela vere sdreges a' bravivoelodok node-didi iniraikee, mad ysdre voi arnageyker:

This dissertation introduces and begins an investigation of an MT model consisting of a novel combination finite-state devices. The model proposed is more flexible than transducer models, giving increased ability to handle word order di ferences between languages, as well as crossing and discontinuous alignments between words. The linked automata MT model consists of a source language automaton, a target language automaton, and an alignment table, a function which probabilistically links sequences of source and target language transitions. It is this augmentation to the finite-state base which gives the linked automata model its flexibility.

Ne elankes ineydamidi mageel wieyn kneides wog n'andredora ys elelang in vaos an gidigrovre MA. Bravivoelodok node-didi mikani arnaelidora slasdem, zoeli elomodeg an bawer, slivi daws slami badredoiel a' slervi ineyn ne raeyngidoraeyn ys MA slasdem. Ranode-didi aeknoreydy inne inddrikdovi vekieyi ddeyr kna ysdre asoeels veryr ekored brakdyang wosan in weelel-eyngerdaas nas weelel-neeirkees gamian. Mad node-dide MA inielokidoraeyn slivi embelaases arnageykereyn an rai rarm ur naaser. Ard areyn ne nideyne ys ddei gevokdy a' aogdelyr kaeybeli slaeyrki nas airged elnageyigi wargeyn vyr mena ys slangeli arnaodora. An dyreki sleyk gevokdy eyes ineyn MA slasdemeyn inne andreges aa kamvani elnageyigi nas arnaelidora mageeleyn anda rae.

Ineyn wieyn gemradrideg, soeyn kaeybelang mikdy ne gonekd nebner edidora ys warg-arger go errekdy vedwere elnageyigdy reyodi gokeyeld, nas bady inggodoraiel kielelregdy rar ne nebner edidora ys inelognmred vedwere slereyrekdy ys warg, an birdokeyelir sai sid inne gokradaneyaey:

the object of study is a two-dimensional information production process, i.e. where one has sources (e.g. journals, authors, words, ...) which produce (or have) items (e.g. respectively articles, publications, words occurring in texts, ...) and in which one considers different functions describing quantitatively the production quantities of the different sources.

All functions can be reduced to one type of function, namely the size-frequency function f: such a function gives, for every n = 1, 2, 3, ..., the number f(n), being the number of sources with n items. This is the framework of study and in this framework we want to explain as many regularities that one encounters in the literature as possible, using the size-frequency function f.

Power Laws in the Information Production Process: Lotkaian Informetrics
Leo Egghe, Elsevier - 2005

Stone Soup Translation: the linked automata model.
Paul C. Davis, The Ohio State University 2002

Towards a uni ed approach to memory- and statistical based machine translation.
Marcu, Daniel. 2001. In Proceedings of the 39th Annual Meeting of the Association for Computational Linguistics, Toulouse, France.

Improved statistical alignment models.
Och, Franz Josef, & Hermann Ney. 2000. In Proceedings of the 38th Annual Meeting of the ACL, Hong Kong, China.

Compositional Translation.
Rosetta, M.T. 1994. Dordrecht, The Netherlands: Kluwer.

Stochastic inversion transduction grammars and bilingual parsing of parallel corpora.
Wu, Dekai. 1997. Computational Linguistics.