|
 |
Building your own corpus | ||
Tags: Comprehensive input | ||
Share with:
| ||
 Language Learning Forum : Learning Techniques, Methods & Strategies |  | |
| Lizzern Diglot Senior Member Norway Joined 5912 days ago 791 posts - 1053 votes  Speaks: Norwegian*, English Studies: Japanese | Message 9 of 27 21 February 2014 at 11:22am | IP Logged | |||||||||||
|
That's a great idea :-) I might start doing something similar for Japanese, but I have no programming skills at all, so I might need to just put my texts in a big Word file to make them searchable. Close enough. Thanks Tommus for the link to WebCorp!
Liz 1 person has voted this message useful | ||||||||||||
| andras_farkas Tetraglot Groupie Hungary Joined 4903 days ago 56 posts - 165 votes Speaks: Hungarian*, Spanish, English, Italian | Message 10 of 27 21 February 2014 at 3:42pm | IP Logged | |||||||||||
|
Interesting. Corpora, especially bilingual and multilingual corpora are kind of my
area, so here are some tips. If you can find multilingual aligned corpora of interest to you, you can get a really nice side-by-side lookup feature. As an example, I grabbed some of the English-French books in my multilingual aligned book collection. It ended up being 45000 sentences, which is a fraction of all the en- fr material I have. I dumped the texts in my lookup software and ran a few sample searches to show you the interface: 65 English sentences contained the word 'anguish': 
In 54 of those 65 sentences, the word was translated as 'angoisse', which shows that the two terms are a good translation for each other. 
The books are here, and the lookup tool (TMLookup) is here. You can also use TMLookup for monolingual texts of course, or, alternatively, you can use apsic xbench, which works largely the same way. Edited by andras_farkas on 21 February 2014 at 3:44pm 7 persons have voted this message useful | ||||||||||||
| Jeffers Senior Member United Kingdom Joined 4912 days ago 2151 posts - 3960 votes Speaks: English* Studies: Hindi, Ancient Greek, French, Sanskrit, German | Message 11 of 27 23 February 2014 at 1:51pm | IP Logged | |||||||||||
You do realize that the English side is the translation, right? This might sound picky to you, but it matters because of what you wrote below:
That is one possible interpretation. The other is that the English translator was lazy and since the words are cognates s/he just used the nearest English equivalent. Cognates are not always the best translation for each other, because the meaning could have shifted in one or both of the languages. For example, you might read, "L'homme est sensible", and be tempted to translate it as "The man is sensible." There is a chance that would be the best translation, but probably not in this case. This type of translation is also poor when the cognate in one language is common, but the cognate in the translation language is not used as much. This seems to be the case with this passage. I am not sure if 'angoisse' is so common in French, but 'anguish' is clearly overused in the English translation, and it seems awkward. My guess is that it has a slightly broader semantic range in French than in English, and doesn't sound awkward in the French. Edited by Jeffers on 23 February 2014 at 1:53pm 2 persons have voted this message useful | ||||||||||||
| andras_farkas Tetraglot Groupie Hungary Joined 4903 days ago 56 posts - 165 votes Speaks: Hungarian*, Spanish, English, Italian | Message 12 of 27 24 February 2014 at 9:35am | IP Logged | |||||||||||
|
While some of your comments are interesting, you miss the point in two ways.
First of all, this was a random sample search to show the way the software works. The content hardly matters. Second, what on earth made you think that "the English side is the translation" in a text collection you've never seen? In some cases, it is, in some cases, it isn't. There are a couple of English novels translated into French in the dataset. E.g. 3 occurrences of the word come from Moll Flanders, and one comes from Robinson Crusoe. In the grand scheme of things, if a word corresponds to X in literary translations in 54 cases out of 65, then it is generally a good idea for language learners to remember that X is a pretty good translation for the word in question. Everybody with an interest in language learning knows that there is no 100% correspondence between words in different languages, so, in almost all cases, any translation is an approximation. The more interesting point here is that collections of aligned texts and software tools of this sort make it really easy to study things like this. You can take a million words of original English text and a million words of original French text and see how often 'anguish' and 'angoisse' occur. Then you can take the opposite language translations of the same texts and compare the word frequencies to see if 'anguish' creeps into English texts as a convenient translation for 'angoisse'. It would be pretty easy to calculate a % value for how often the similarity of the cognates leads the translator astray. You could even use brute force to automatically find anomalies. I.e. run searches on the 10,000 most frequent words in texts translated into English, then run searches on the same words in original English texts and see where the biggest discrepancies are and what they may be due to. If there is enough data, it wouldn't be too hard to build up profiles of individual translators and see how their language background affects their work. In fact I expect references to farkastranslations.com to show up in linguistics/translation studies thesis papers a few years from now. Edited by andras_farkas on 24 February 2014 at 9:47am 5 persons have voted this message useful | ||||||||||||
| Jeffers Senior Member United Kingdom Joined 4912 days ago 2151 posts - 3960 votes Speaks: English* Studies: Hindi, Ancient Greek, French, Sanskrit, German | Message 13 of 27 24 February 2014 at 12:53pm | IP Logged | |||||||||||
I would like to say the content always matters, but I take your point. Nevertheless, you drew conclusions based on this sample. Perhaps this means it is a good example of how the software could be misused?
It seems pretty obvious that the original is the French side, since all but one of the character names are French (and the one English name is identical in French so it could be French as well). Computer tools are wonderful things, I agree. I am a teacher of ICT and Computing, so I value them greatly. But it is also too easy to make mistakes based on computer tools. For example, if you unwittingly search a text translated into English, you might draw conclusions about English word frequency or English syntax which may or may not be correct. The problem is that a computer tool can't apply common sense, and so the user needs to be smarter than the system in order to avoid false conclusions. You mention the possibility of searching original English text separately from original French text. That sort of distinction would go a long way towards avoiding mistakes about things such as word frequency. 1 person has voted this message useful | ||||||||||||
| andras_farkas Tetraglot Groupie Hungary Joined 4903 days ago 56 posts - 165 votes Speaks: Hungarian*, Spanish, English, Italian | Message 14 of 27 24 February 2014 at 1:27pm | IP Logged | |||||||||||
No, I still think it's an example of how the software could be used.
Talking about avoiding false conclusions, all but one of the character names are French in the dozen sentences you can see in the screenshot. As I said, there are 65 hits in a random collection of texts totalling about 45,000 sentences. As it happens, most of it is made up of French originals translated into English, but not all. If one were to use this tool for scientific research, one would obviously put the various types of texts (English original, French original) into separate databases. For language learning, this is hardly a vital issue. One tries to read texts that were originally written in the target language, but it's not a hard and fast rule. Edited by andras_farkas on 24 February 2014 at 1:31pm 4 persons have voted this message useful | ||||||||||||
| lingoleng Senior Member Germany Joined 5301 days ago 605 posts - 1290 votes | Message 15 of 27 24 February 2014 at 4:32pm | IP Logged | |||||||||||
If I am not completely mistaken Andras is a professional translator and interpreter. Kind of funny that you feel the urge to lecture him about some elementary basics ... 1 person has voted this message useful | ||||||||||||
| Jeffers Senior Member United Kingdom Joined 4912 days ago 2151 posts - 3960 votes Speaks: English* Studies: Hindi, Ancient Greek, French, Sanskrit, German | Message 16 of 27 24 February 2014 at 5:53pm | IP Logged | |||||||||||
|
Good answer: back off, we're professionals.
I'm sorry, but if a "professional" can't accept criticism of their work, then I have no respect for them as a professional. A true professional takes on board criticism and improves. Being defensive simply compounds errors. And professionals closing rank against any criticism, well that just makes the profession look weak. I wasn't going to write anything further, because andras_farkas didn't want to hear it. However, I was disgusted at the attempt to shut down criticism based on credentials. Trust me, I will leave the rest of this discussion to the "professionals". Good day to you sir. 2 persons have voted this message useful |
|  | 
Printable version
|
| Copyright 2024 FX Micheloud - All rights reserved No part of this website may be copied by any means without my written authorization. |
