Talk:History of machine translation
 | Translation studies C‑class | |||||||||
| ||||||||||
 | Computing C‑class | |||||||||
| ||||||||||
This page
This page is largely paraphrased from [1]. I've asked a few people to look over it before I pasted it in, and the majority said it didn't qualify as {{copyvio}} because nothing is duplicated, one person said it was "close to the wind". I'd welcome improvements. If anyone thinks it is a copyvio, please don't hesitate to put the notice on! :) - FrancisTyers 15:04, 25 April 2006 (UTC)
The beginning
In the part where you state that the beginning of machine translation lies in the universal language ideals of likes of Descartes and Leibniz, I'd dare to disagree. If the status as a precursor of modern automated translation is given to concepts of universal code between languages wouldn't it be more rational to give the credit to universal grammarians? Hans Aarsleff suggests on his Chomsky criticism that first such grammers would have been made during the late medieval times. After all, in philosophy the idea of natural origin of signs would likely date back to ancient Greek in which case the nature would have been the first tertium comparationis that later on got replaced by mind in cartesian philosophy.
If on the other hand actual machines are treated then e.g. Schubert (2006; 1140)writes in Eroms, Hans Werner (ed.) Dependenz und Valenz: Ein Internationales Handbuch Zur Zeitgenossischen Forschung (2006)that,
Die Idee der Automatisierung des Übersetzens bringt etwa vom 18. Jahrhundert an zahlreiche Pläne und Entwürfe übersetzender Maschinen hervor.
This would suggest that already at 18th century onwards there have been special machines reserved for this purpose akin to Pascaline.
--78.27.78.126 (talk) 15:08, 2 November 2009 (UTC)
This should help with the referenced quote
FrancisTyers, sorry you had such a hard time finding this part of the history of machine translation, here is the full quote. At your service, let me know if I can help in the future. dbp653
Start of quote:
BETTER TRANSLATON FOR BETTER COMMUNICATION A survey of the translation market, present and future, prepared for the Commission of the European Communities, Directorate-General Information Market and Innovation by Bureau Marcel van Dijk, Brussels PA Conseiller de Direction, Paris
Authors G. Van Slype (Bureau Marcel van Dijk) J. F. Guinet (PA) F. Seitz (PA) E. Benegam (PACTEL)
Published for the COMMISSION OF THE EUROPEAN COMMUNITIES by PERGAMON PRESS OXFORD . NEW YORK . TORONTO . SYSNEY . PARIS . FRANKFURT
(Next Page)
Published for the Commission of the European Communities, Directorate-General Information Market and Innovation -- Scientific and Technical Communication, Luxembourg
Copyright 1983 ECSC, EEC EAEC Luxembourg
ISBN 0-08-03053402 EUR 7720EN
Chapter 5
Machine Translation
118
5.3. History of machine translation
119
5.31. Before 1954: the precursors.
5.32. From 1954 to 1966: the pioneers.
Around 1964, the American National Science Foundation noted that the large investments in MT had not achieved any practical result and did not seem likely to do so in the foreseeable future. It therefore asked the National Academy of Sciences to set up a working party to advise the institutions which were giving subsidies: the National Science Foundation, but also the Department of Defense and the Central Intelligence Agency (CIA), on the value of the policies they were following. This group known as the ALPAC Committee (Automatic Language Processing Advisory Committee) was set up in 1964. It carried out a number of interviews and commissioned a number of studies Its Report, published in 1966, concluded that: - MT is slower, less exact and twice as expensive as HT.
121
The Report made the following recommendations: - The subsidies to (applied) research on MT should be stopped;
5.33. From 1966 to 1978: the crossing of the desert.
The report of the ALPAC committee led to the virtually complete withdrawal of subsidies to research into MT in the United States and a number of other countries.
in the United States: the BYUALP project (Brigham Young University Automated Language Processing) directed by Eldon Lytle and intended to translate Mormon "proselytic kits" from English into French, Spanish, German and Portuguese. This project was launched in 1971 and dropped in 1979 (Melby-1979)
124
- by the virtually complete abandonment of applied research on MT in the United States and, on the contrary, its development in Europe and Asia. - by the hesitant appearance of a few operational systems which only found limited application in very specific contexts: obviously, there was no longer a market for MT.
5.34. 1978: hope renewed.
From about 1978 onwards, a new wind began to blow in the world of MT.
125
- a new translation system was developed in the United States and marketed simultaneously in the United States and in Europe by the Weidner company. The unusual formula proposed (sale at a fixed price of a complete system including a minicomputer, translation terminals and software) seems to have had some success among translation services and agencies.
128
- WEIDNER.
Following an unpublished technico-economic evaluation in 1979, the Mormon community of the Brigham Young University, in Utah, deciede to scuttle its own system, which it felt was based on too academic an approach, and to invest five million dollars in the undertaking which the Weidner brothers had just established, also in Utah: Weidner Communications Inc. Conceived from the start as a text processing system linked to an interactive machine translation system, with dictionary updating by the translator, the Weidner system made an explosive start (Haymond - 1980), thanks to the establishment of a large team of approximately 50 persons:
129
A daring marketing policy has been developed, which consists in selling the translation systems on a turnkey basis at a fixed price by language pair, including a Digital Equipment minicomputer, terminals, translation software, the text processing software and a current vocabulary dictionary comprising approximately 8000 roots and 3000 expressions. A maintenance contract provides of the delivery of improved versions of the software every six months. The translation speed, in batch processing is around 16000 words/hour
130
- The WEIDNER system was rapidly evaluated by a translator of the CEC interested in MT (Pigott - 1980): his conclusion was that the quality of the Weidner Spanish-English system was comparable to that of Systran French-English.
end of quote.
I would also highly recommend this material for the History of Machine Translation: Tim Johnson: Natural Language Computing: The commercial applications, London 1985: Published by Ovum Ltd, 1985, “the situation is beginning to change” (computers processing human language) he attributes to “Weidner.”
It is interesting that all this information is out there for all to see but I am being harassed for sharing it on Wikipedia.
- You are not being harassed; you are being asked to obey WP:COI guidelines avoid direct editing of articles where you have a clear self-described conflict of interest (see WP:COI/N#Bruce Wydner). It would help if you grasped the consensus - see Wikipedia:Articles for deletion/Language technology - that you are approaching this subject in a way that is not acceptable here. Please also stop posting these lengthy duplicated justifications; it is viewed as disruptive. Gordonofcartoon 03:08, 27 June 2007 (UTC)
Referencing
I've given this article a bit of a copyedit, but in the process I've sprinkled cn tags through it. I'm noting here that I intend to read up on the subject and add some referencing, and perhaps a bit more detail. If anyone could suggest a good book covering a large fraction of the history mentioned, could they please post it at my talk page? Thanks. GoldenRing (talk) 04:00, 27 January 2015 (UTC)
2010's
Has there been enough developments in the past 7 years to warrant the inclusion of a new decade section?
I think that Google Translate [1] and Facebook's [2] use of Neural Machine Translations are pretty significant advances that have only been developed in the past two years, but there's no use of the phrase Neural Machine Translation (or similar) anywhere in the article.