| By Kevin Hoffman | Article Rating: |
|
| September 17, 2009 08:33 AM EDT | Reads: |
672 |
Before I get into the main part of the blog post, I want to preface this whole thing by saying that I am well aware of the problems of doing language translation by a machine. It's an inherently difficult problem.
My issue here isn't so much with Google Translate as it is with the public perception of Google. I've read a few reviews of Google Translate and seen some blog posts. The general consensus is that because Google built it, its awesome and flawless. People think that Google is building this massive, soon-to-become-self-aware giant brain in the cloud that is sucking intelligence out of search patterns to be used for some world dominating purpose in the future. This is rubbish because we know Google doesn't start Judgement Day, Cyberdyne does. Sheesh, anybody knows that.
So, what's different about Google Translate than other previous attempts at machine translation? The biggest is that Google is using collective intelligence like it uses on its search engine to cull human-supplied translation bits and add those to its existing translation rules. The theory is that it makes for a better engine and should make the translations more in line with what a human would expect.
So how does it fare? You tell me:
| My Input | Google's Output | Actual Meaning |
| 内閣は倒れるだろうということだ | Cabinet and they will fall | They say that the cabinet (govt) will fall |
| その先生は病人だということが分かった | The teacher found that it's sick | The teacher turned out to be sick |
| 私達は彼の死を悲しんだ | We mourned the death of his | We mourned his death |
| 彼の殺人の動機は何だ |
What's his motive for murder
|
What's his motive for murder?
|
So this is actually pretty enlightening. I have to admit that I was really skeptical when I tried it out. Just like every other machine sentence translator on the planet, if you attempt to use a Google translation to glean meaning from a web page or from a paragraph or even a single sentence, you're probably screwed. The first three sentences I have in the chart show that while verbs and nouns are translated properly, you actually lose a lot of meaning and context in translation. For example, the second sentence - relying on the Google translation I don't know if the teacher is sick or something else - but it's pretty damn close to accurate. The second to last sentence is almost spot on and finally, the last sentence is a perfect translation.
So the verdict is this: the current state of machine language translation sucks and will continue to suck for some time. However, I now know that if I want to get a rough stab at a translation of something I am going to try Google Translate first. Among all the other engines I could find, Google's was far and away the best.
For example, when I plug in the Japanese for "what's his motive for murder?" Babelfish spits out the following: Motive of at that homicide is what.
Someday this won't matter and we'll all have universal translators in our ears. It also doesn't matter much because, as Hollywood has shown us time and time again, all aliens from far and distant worlds speak English.
Read the original blog entry...
Published September 17, 2009 Reads 672
Copyright © 2009 SYS-CON Media, Inc. — All Rights Reserved.
Syndicated stories and blog feeds, all rights reserved by the author.
More Stories By Kevin Hoffman
Kevin Hoffman, editor-in-chief of SYS-CON's iPhone Developer's Journal, has been programming since he was 10 and has written everything from DOS shareware to n-tier, enterprise web applications in VB, C++, Delphi, and C. Hoffman is coauthor of Professional .NET Framework (Wrox Press) and co-author with Robert Foster of Microsoft SharePoint 2007 Development Unleashed. He authors The .NET Addict's Blog at .NET Developer's Journal.
- Kindle 2 vs Nook
- Installing Geneva Beta 2 on Windows 7
- Binary Serialization and Azure Web Applications
- Get Your Red Hot VS2010 Beta 2
- Templated Helpers in ASP.NET MVC 2 (VS2010 Beta 2 Version)
- LINQ to SQL and Entity Framework on top of SQL Azure
- Working with Table Storage on the Windows Azure
- ADO.NET Data Services Projections Makes Sliced Bread Jealous
- Creating and Manipulating Your SQL Azure Database
- Setting up an ASP.NET MVC 2 Application for Windows Azure
- Breaking Changes for .NET Services in Azure
- Windows Identity Foundation (WIF) Release Candidate Is Out
- Kindle 2 vs Nook
- The Difference Between Web Hosting and Cloud Computing
- ASP.NET Membership Provider in the Cloud
- Installing Geneva Beta 2 on Windows 7
- Binary Serialization and Azure Web Applications
- Get Your Red Hot VS2010 Beta 2
- Templated Helpers in ASP.NET MVC 2 (VS2010 Beta 2 Version)
- LINQ to SQL and Entity Framework on top of SQL Azure
- Using ASP.NET MVC Action Filters to Declare Reference Data for Views
- Creating Correlated Workflow Services in WF4 / .NET4 : Part 1
- Working with Table Storage on the Windows Azure
- ADO.NET Data Services Projections Makes Sliced Bread Jealous
- Want to Learn How to Write iPhone Applications?
- iPhone Will Make Mobile AJAX and Web 2.0 Happen
- Will Silverlight Be DOA?
- Why Build Applications for the iPhone and iPod Touch?
- Silverlight 2 - Adobe Flex Killer Is on Its Way!
- Why Is iPhone Better? Here's My Story...
- Silverlight and Astoria - First Impressions
- iPhone Developer Summit 2008 East
- Is the Silverlight Adoption Rate Artificially Inflated?
- iPhone with High-Speed G3 Support at Macworld
- Will Google's Android Sink or Swim?
- iPhone Price Cut? Here is My Objective View on This!


























