Google recently launched the Translator Toolkit, which helps users translate and edit documents or web pages from one language to another. This is industrial scale crowd sourcing and machine learning at it's finest and is a rare win-win-win situation:
- Users get tools that help make their job easier. With a few click a document gets processed and translated automatically. The user then has to proofread, edit, and publish the results, but they get access to tools like dictionaries, glossaries, and previous translations that may be related.
- Partner websites like Wikipedia and Knol will get better translations for articles because they are edited and approved by humans.
- Google gets to train it's translation algorithms using data from the human edits to improve it's accuracy. I took a basic natural language processing class a few years back, and getting access to mistakes and their correlated corrections is invaluable for refining and improving these types of algorithms.
If you are interested check out http://translate.google.com/toolkit or just watch the video:
No comments:
Post a Comment