TurtleWeb Forums -- Supporting the TurtleWeb Community   Search   Recent Topics     Member Listing    Back to home page   Login / Register

Forum Index Profile for :: atorok Avatar All about atorok Ranking: Karma: Registration date:  09/02/2023 02:09:49 Number of messages posted:  No posted messages available Created topics: No topic created Website:  http://keycodesoftware.com/ Occupation:  ??????-?????????????? Interests: ???????? ???????, xml Biography: On wednesday, a group of microsoft researchers announced the formation of the first machine translation system capable of translating news articles from chinese to english with as much accuracy as an employee. The company proves that it has repeatedly tested the system on a sample of about two thousand offers from various online newspapers, comparing the result with the language support of any of us in the game, and even hiring external bilingual consultants to internally check the accuracy of the machine. A sample set called newstest2017 was presented last fall at the wmt17 research conference. It's amazing how quickly researchers have been able to reach this milestone, let's not forget that machine translation is a problem that people have been trying to solve throughoutduringduringduringduringduringduringduringduringduringduringduringduringduringin duringduringduringduringduringduringduringduringduringduringduringduringduring duringduringduringduringduringduring decades. Mn some even thought that the goal of human parity would never be realized, says microsoft. “Achieving human matching in a machine translation task is a delight for everyone,” said xuedong huang, technical officer in charge of microsoft's speech, natural language, and machine translation efforts at microsoft. “We just didn't realize it was possible to achieve this so soon.” Making a machine understand language on such a scale is a lot more boring than speech recognition—something before the number of advances in recent years. Advances in ai and speech recognition have allowed voice assistants to permeate our tablets and our homes, where they help consumers with everyday computing tasks, smart home device control, and events and fun activities. But the problem is that the machine translation of a web page or a news article still often renders that same difficult-to-understand set of words that, at best, gives you a generalized knowledge of everything that is being talked about, but is almost impossible to understand with any deep understanding. To really find out what is being said in longer articles, you need human knowledge. But even all human translators can translate a sentence a bit clumsily. In different ways, and none of them are wrong. "Machine translation is much more difficult than the simple task of pattern recognition," said ming zhou, assistant managing director of microsoft research asia and head of natural language. The processing team that worked on the project. “People are ready to use different words to express the same thing, but you can’t say in 7-24 mode which of the gambling establishments is better.” Recent breakthroughs in the field of ai. Microsoft also notes that it helped researchers achieve this milestone. Deep neural networks, an ai learning method. Systems have allowed researchers to create smoother, more natural-sounding translations that take into account a much more diametrical context than early approaches called statistical machine translation. Microsoft researchers have also added their own learning options to the system to support its accuracy is an accessory that they equate to the way people check their profession again to make sure it's correct. Researchers said they used methods including double learning to check translations; deliberative online, for repeating translations and refining them, and new techniques, including collaborative learning to greatly improve translation systems from english to chinese and from chinese to english; and regularization of conventions, which can generate translations by reading sentences side-by-side and right-to-left. Zhou said the methods used to reach the milestone will not be limited to machine translation. “This is a niche, and here machine translation research fits into the entire field of ai research,” zhou said. Also, it can offer more accurate and lively translations to other languages to come. Researchers warn that the system has not yet been tested on real-time incidents, and there are challenges still ahead of whether the technology can be commercialized in microsoft products. But you can play around with a different translation system here on the microsoft portal: https://translator.Microsoft.Com/neural.(This is unlikely to be a production system, and sometimes the camera may be sluggish, the site warns.) The system will display an ad in chinese (simplified), which is then translated in two ways: improved translation on the right to show improvements . Google researchers are also working on machine translation, even with their own machine learning technology for chinese-english queries, which makes heavy use of neural networks. These advances are already being applied to transform google's consumer-facing products, such as google translate and its integration into google search. For information about https://keycodesoftware.com/, please visit our webpage. My Bookmarks: There are no bookmark entries for this user. Contact atorok Private Message: MSN Messenger: spectacularpoke Yahoo Messenger: spectacularpoke ICQ UIN:

Powered by JForum 2.1.7 © JForum Team