Talk:Translation memory

From Wikipedia, the free encyclopedia

Translation memories generated by MT programs such as Trados, TM2, Transit and the like, as I understand, work with identifying a number of (generally 7)(semi)identical words in a segment to be translated (L1) and a segment that was to be (L2m) and actually translated earlier (L2n). On the basis of that match they will offer the sentence in L2 for checking whether it corresponds to the original and earlier input in L1n and therefore its reuse is justified. But, the syntagmatical and paradigmatical features of English (Indo-Euuropean languages) and other languages such as Hungarian are so much apart that this mechanical match does not work reliably and efficiently, especially backward, from L2 to L1. As long as the intellegence built in MT is focused around AI, including matching and fuzzy matching you cannot expect more than fuzy hits, just as in the results of machine/mechanical indexing used in most of the Internet search operations, engines, etc. Apogr 12:21, 15 Feb 2004 (UTC)

Does anyone new free translation memory ? -- Youssefsan

This tool is all right as long as you work with two indo-european languages. But it is a nonsense, if you are required to use Trados for instance to translate from English into Hungarian and vice versa. A better approach to the problem is when you do more pre-procesing and less post-processing after translation and hire highly skilled translators as opposed to employing novices who are fed the machine translated garbage for "editing". See the practice of the translation center(s) of the European Union

The idea of translation memory comes from the false assumption of making translation work more efficient by re-using former translations in a particular subject where the vocabulary is almost controlled or recurrent. Translators, as a result have become more involved in the issues of editing, searching and formatting texts, aligning their segments, phrases, etc. that drastically decreases their productivity and lowers the demand on their skills.

Since most of the translation done from English into other languages (including Hungarian) consists of inventing new words, explaining terms that are kept in their original forms, etc. because of the very nature of the content of the works to be translated, translation memories just as other software developped in the English speaking world are a nuisance to non-native-English speaking translators who try to comply with the rules of their native tongues in translating non-repetitive texts.

In Hungarian for the past twenty years not only the vocabulary, the signs, company names, etc., but e.g. the ways nouns and adjectives are put together, verbs are inflected, addresses, dates and names are written, etc. have changed for the less-complying and new varieties of pidgin and lingo are to emerge. See also translation, translation problems Apogr 09:15, 9 Jan 2004 (UTC) more to come, if solicited Apogr 09:15, 9 Jan 2004 (UTC)

This information is disputable: "As stated previously, translation memory managers may not be suitable for text that lacks internal repetition or which does not contain unchanged portions between revisions. Technical text is generally best suited for translation memory, while marketing or creative texts will be less suitable." Translation memory managers (Cat tools would be a better term) are suitable for any translation work. They prevent skipping text, usually have a glossary manager built-in and other features that help translators to work faster.

I added the "Second generation translation memories" paragraph in the "History" section. I own the copyright of this text extracted from the Similis web site. If someone needs a proof of this, thanks to give me a procedure. EMLM 11:39, 18 August 2006 (UTC)

On the whole, both of the points of view are correct. What regards the first one. Using CAT tools in translation is aimed at speed and general productivity increase and applies to most languages supported by information systems. Due to its core principle, translation and localization (if that is the case) is done by the linguist (translator) and is independent on the tool used. Though high overall leverage and fine match rates are harder to obtain when working with texts of non-roman language origin, such as cyrilic, Asian etc. Danz0 13:55, 29 September 2006 (UTC)

[edit] Free versus commercial

I dislike the distinction "free" versus "commercial". Free tools are not necessarily non-commercial tools. There has got to be some other words one can use. Or, here's a thought... why make the distinction at all? Why not add "(freeware)" or "(free version available)" or "(GPL)" or "(time-unlimited shareware)" or similar next to the ones in the list? leuce 10:13, 18 December 2006 (UTC)