Showing posts with label machine translation. Show all posts
Showing posts with label machine translation. Show all posts
Thursday, May 22, 2014
When #Google is my friend; translating from #Marathi
I am adding dates of death to many people who died in 2014. Included in my workflow are people with an article on the Marathi Wikipedia. It is well known that translating into Marathi for articles is not great. For my purpose; identifying a date of death and a date of birth it is good enough :)
Thanks,
GerardM
Tuesday, April 15, 2014
#Wikidata - The dead at the #Arabic #Wikipedia
As the #quality of Wikidata can be measured, it is important to be inclusive when what is measured are the people who died in 2014. People die in every country like this gentleman from Saudi Arabia.
According to the article, he died Thursday, 12 May 1435 AH, that is the same as 13 March 2014. Google translate transliterates the title as "Zaid bin Mohammed portal". That is enough reason not to use the transliteration as the title for the item.
When you read the article it is about Mr bin Mohammed and does not give the impression of a portal page.. That is for someone who understands the Arabic Wikipedia to fix.
Thanks,
GerardM
According to the article, he died Thursday, 12 May 1435 AH, that is the same as 13 March 2014. Google translate transliterates the title as "Zaid bin Mohammed portal". That is enough reason not to use the transliteration as the title for the item.
When you read the article it is about Mr bin Mohammed and does not give the impression of a portal page.. That is for someone who understands the Arabic Wikipedia to fix.
Thanks,
GerardM
Wednesday, February 15, 2012
#Google you do not need to guess
#Wikipedia supports more then 280 different projects and each and every one is in one language. In the name of the project you find an indication of what language it is, in the meta-data you find what language it is.
As a consequence it should be obvious that the tk.wikipedia.org, the one in the Turkmen language is not written in Turkish. Your software does not need to guess what language it is because we make it plain. It says somewhere in the code:
Thanks,
GerardM
As a consequence it should be obvious that the tk.wikipedia.org, the one in the Turkmen language is not written in Turkish. Your software does not need to guess what language it is because we make it plain. It says somewhere in the code:
<div lang="tk" dir="ltr" class="mw-content-ltr"><table border="0" cellspacing="5" cellpadding="0">In my understanding it means that the content that follows is in the "tk" language. Yes, I will continue to report to you that you reported an incorrect language. The only thing, you do not need to guess what language it is, we do indicate it properly.
Thanks,
GerardM
Sunday, March 06, 2011
Are you ready for the new #translatewiki.net server ?
With a new server running translatewiki.net, we are again ready for old and new business. This is the time to hammer our server by localising some of the best open source software around.
With more hardware resources we will provide our community with a better service. Whenever you have a spare moment, we are quite happy to make the most of your time and make you as productive as we possibly can.
Our aim is to make your work as easy and relevant as we can make it. LocalisationUpdate ensures that your MediaWiki localisations have the fastest turn around time into production. Narayam helps you to enter the characters of the script for your language from a standard keyboard. Machine translation engines are available to make suggestions to you.
With you localising software, we do what we can to utilise the technology that makes your time with us effective and ensure that your work makes a difference.
Thanks,
GerardM
With more hardware resources we will provide our community with a better service. Whenever you have a spare moment, we are quite happy to make the most of your time and make you as productive as we possibly can.
Our aim is to make your work as easy and relevant as we can make it. LocalisationUpdate ensures that your MediaWiki localisations have the fastest turn around time into production. Narayam helps you to enter the characters of the script for your language from a standard keyboard. Machine translation engines are available to make suggestions to you.
With you localising software, we do what we can to utilise the technology that makes your time with us effective and ensure that your work makes a difference.
Thanks,
GerardM
Thursday, August 05, 2010
The limits of #Google #Translate and other translator tools
One of my #Wikipedia friends is Bellayet. He has represented the Bengali Wikipedia at Wikimania several times, I have corresponded with him on several occasions; he is one of the good guys.
Bellayet blogs and uses the Wordpress software. I am happy to notice that its interface looks Bengali to me. I asked Bellayet if he could add something like Google translate to his blog and learned that he could not because Bangla is not (yet) supported.
There is a lot of talk about pushing translations using tools like Google translate. The reality is that Bangla, a language spoken by some 250,000,000 people is not supported to the extend that something can be pushed out. Something that is enough to help me understand what the subject is, maybe even understand somewhat what is being said.
Google also supports a translation tool that includes the use of translation memories and machine translation. It is vital to understand the tool, what it does and what is expected of you the translator.
This toolkit gives you the kind of tools that are also used by professional translators. It makes sense to appreciate what it is a translator does and does not do because once you do you have realistic expectations.
Typically a translator only translates into his or her mother tongue. They use tools to be as efficient as possible. Many translators translate the same kind of text really often, think travel documents, health documents etc. They have a fixed vocabulary and are in a standardised format. For such activities translation memories are a godsend.
Another group of documents, for instance Wikipedia articles, are hard on translation memories. The subject is singular and therefore there is often not that much to base a proposed translation on. This is where machine translation is of use. It has a dictionary component and it has a morphology component. Together with some grammar rules they provide the best they can.
When you as a translator provide a properly written literal translation, the translation can be used to enrich a translation memory. However, many translators have to provide a text that fits well with an audience, a different culture. In essence a text is substantially re-written to fit its target.
Another task often required of a translator is formatting. In a wiki environment, this would include wiki-links, templates and pictures. All of this is what is part of providing an end product that is acceptable to its public. When professional translators are hired to translate, it makes sense to hire Wikipedians to finish the translation by providing the formatting.
Thanks,
GerardM
Bellayet blogs and uses the Wordpress software. I am happy to notice that its interface looks Bengali to me. I asked Bellayet if he could add something like Google translate to his blog and learned that he could not because Bangla is not (yet) supported.
There is a lot of talk about pushing translations using tools like Google translate. The reality is that Bangla, a language spoken by some 250,000,000 people is not supported to the extend that something can be pushed out. Something that is enough to help me understand what the subject is, maybe even understand somewhat what is being said.
Google also supports a translation tool that includes the use of translation memories and machine translation. It is vital to understand the tool, what it does and what is expected of you the translator.
This toolkit gives you the kind of tools that are also used by professional translators. It makes sense to appreciate what it is a translator does and does not do because once you do you have realistic expectations.
Typically a translator only translates into his or her mother tongue. They use tools to be as efficient as possible. Many translators translate the same kind of text really often, think travel documents, health documents etc. They have a fixed vocabulary and are in a standardised format. For such activities translation memories are a godsend.
Another group of documents, for instance Wikipedia articles, are hard on translation memories. The subject is singular and therefore there is often not that much to base a proposed translation on. This is where machine translation is of use. It has a dictionary component and it has a morphology component. Together with some grammar rules they provide the best they can.
When you as a translator provide a properly written literal translation, the translation can be used to enrich a translation memory. However, many translators have to provide a text that fits well with an audience, a different culture. In essence a text is substantially re-written to fit its target.
Another task often required of a translator is formatting. In a wiki environment, this would include wiki-links, templates and pictures. All of this is what is part of providing an end product that is acceptable to its public. When professional translators are hired to translate, it makes sense to hire Wikipedians to finish the translation by providing the formatting.
Thanks,
GerardM
Wednesday, July 14, 2010
#Google #translate use at #Wikimania
At the dorm where I slept, a nice Polish lady manned the reception. She was nice and she spoke Polish. The Wikimanians did not. So communication was hard until.. until one had the great idea to use translate for communication.
It eased a lot of nerves most importantly of the lady who now could get most of her message out. She wanted to communicate and we wanted to understand, translate provided the interface.
Thanks,
GerardM
It eased a lot of nerves most importantly of the lady who now could get most of her message out. She wanted to communicate and we wanted to understand, translate provided the interface.
Thanks,
GerardM
Friday, June 19, 2009
Wikiblame is now supported by translatewiki.net
Wikiblame is now supported by translatewiki.net. This is cool, when you check out the project page you find that languages that have localised the software, are not supported in the software yet.This is quite something because this may be the first project where the localisation may drive what languages are available for this project.
Thanks,
GerardM
Tuesday, February 17, 2009
Apertium, or benefitting from computational linguistics
There are Wikipedias in many languages. Some languages are written in multiple scripts. And as long as one script has sufficient information it is possible to convert it into another script. When this can be done with a program, it prevents an awful lot of work.
Several languages have a Wikipedia with multiple scripts. For some like Chinese, there is software that converts from one script in another. For other languages it would be nice to have this as well. Fiji Hindi is one such; it is written in the Latin and the Devanagari script. A lot of work has already gone in the Latin localisation at translatewiki.net, and it would be really beneficial when we could automate correctly in the Devanagari script.
I had a talk with my friend Francis Tyers and he is quite willing to help with this. I know Francis from his work on the Apertium translation engine and conversions like this may be doable. Francis is happy to look into the possibility of such a conversion. The only thing we have to solve is how do we get the Latin text out and back into the message file again.
We are looking into the language part of this, now we need someone who can help us with the MediaWiki side of it.. Can you help?
Thanks,
GerardM
Several languages have a Wikipedia with multiple scripts. For some like Chinese, there is software that converts from one script in another. For other languages it would be nice to have this as well. Fiji Hindi is one such; it is written in the Latin and the Devanagari script. A lot of work has already gone in the Latin localisation at translatewiki.net, and it would be really beneficial when we could automate correctly in the Devanagari script.
I had a talk with my friend Francis Tyers and he is quite willing to help with this. I know Francis from his work on the Apertium translation engine and conversions like this may be doable. Francis is happy to look into the possibility of such a conversion. The only thing we have to solve is how do we get the Latin text out and back into the message file again.
We are looking into the language part of this, now we need someone who can help us with the MediaWiki side of it.. Can you help?
Thanks,
GerardM
Wednesday, January 17, 2007
Standards .. we need them
I had another conversation with someone who has a big interest in the Apertium machine translation software. OmegaWiki has something to offer to offer to tools like Apertium; it is a database, it intends to have all words of all languages and last but not least it's data will be available under a Free license.
To be of relevance to tools like Apertium we need to have conjugations and inflections. This is something that we have always planned to include. The new thing for me is that it is desired to have morphological information as well. With the current costs of computer hardware, including the data is not that much of an issue; as it is only text this will not really amount to a big increase in hardware costs.
I then asked the question; is there a standard format for morphological data.. Apparently there is no standard for it. So what to do when there is no standard. You can select one way of showing morphological data amd this will piss people off or you can create a new standard and piss off every one.
The good news is that we first have to deal with the things that come first; being able to include conjugations and inflections. So there is some time before it actually becomes relevant. The thing is it will, and it is likely that we will do something. When we select one way, I hope it will be flexible enough to allow for conversion to other formats as well.
Thanks,
GerardM
To be of relevance to tools like Apertium we need to have conjugations and inflections. This is something that we have always planned to include. The new thing for me is that it is desired to have morphological information as well. With the current costs of computer hardware, including the data is not that much of an issue; as it is only text this will not really amount to a big increase in hardware costs.
I then asked the question; is there a standard format for morphological data.. Apparently there is no standard for it. So what to do when there is no standard. You can select one way of showing morphological data amd this will piss people off or you can create a new standard and piss off every one.
The good news is that we first have to deal with the things that come first; being able to include conjugations and inflections. So there is some time before it actually becomes relevant. The thing is it will, and it is likely that we will do something. When we select one way, I hope it will be flexible enough to allow for conversion to other formats as well.
Thanks,
GerardM
Subscribe to:
Posts (Atom)






