Mandarin Within Reach

Amazing: Microsoft turns spoken English into spoken Mandarin – in the same voice

language translation that goes far past what we thought was currently possible...... has taken in oodles of data, and can thus have that translated Mandarin spoken. And the final kicker: he has fed the system an hour’s worth of his voice, and thus the software will speak in Mandarin, using his own tones. ..... Just over two years ago, researchers at Microsoft Research and the University of Toronto made another breakthrough. By using a technique called Deep Neural Networks, which is patterned after human brain behavior, researchers were able to train more discriminative and better speech recognizers than previous methods. ..... rather than having one word in 4 or 5 incorrect, now the error rate is one word in 7 or 8

This is incredible.

Related articles

• Amazing: Microsoft turns spoken English into spoken Mandarin -- in the same voice
• Microsoft demos real-time English to Chinese translation
• Must-See Video: Real-Time English-to-Mandarin Speech Translation via Microsoft Research
• Microsoft: Star Trek's universal translator not that far off
• Microsoft demonstrates "breakthrough" speech translation technology
• Microsoft's universal translator converts your voice into another language
• 4 Essential Qualities of Successful Mandarin Chinese Learning (Part 3 of 4)
• Will the Cairns Post promote SEXY MANDARIN might sell more newspapers and help the new tourists from China feel at home - - - - What is SexyMandarin? - SexyMandarin
• Pretty Amazing Machine Translation Demonstration
• Rumor: Siri to add Japanese, Mandarin, Russian next month

"Can You Understand This?"

Image by tonynetone via FlickrRadio Nepal would serve the news in Nepali at seven, morning and evening, and the news in English an hour later at eight. This was during the days of the autocratic monarchy. And so there was much state propaganda. I much preferred listening to the BBC. In English, of course.

Of course no one in my village listened to the news in English. The smart ones listened to the BBC Hindi service.

But then there was always some smart alec who would turn the radio on for the eight o'clock news in English.

"Can you understand this?"

"Yes."

Brazil: Historically Speaking

Brazil is not a Spanish speaking country. The leading country in South America is not a Spanish speaking country. The British left, but the English language in India just grew and grew and grew. Brazil used to be a colony of Portugal. Just like India basically took over the English language - there are way many more English speakers in India than in England, way, way more - Brazil has taken over the Portuguese language.

Search: Much Is Lacking

Search as is is not good enough. If you can't find what you are looking for, as often happens. The user is not in control enough.

Search is especially poor on the multimedia front. You have to find a particular site, and then search around inside that site, often you have to be a registered user to make use of the stuff offering.

There is no one place where you can go to search through all photos online, no one place to look for audio and video clips, all of them, no way you can zero in like you might zero in onto a map, zoom in, zoom out.

It would be nice to be able to zero in on content providers based on their geographical location. That would really help.

There is not enough standardization. All of the web is not being shown as one search space. Participating sites would benefit. If you make it easier for users to find you, that is good business, right?

Ad offerings will have to get creative if the ad model is to be the primary revenue generator. Tiny ads. This photo brought to you by Coke. Can you say that in an inch in length, and 1/10th that in width, maybe in less? Of course, you can see the Coke logo which, if you click on takes you to the Coke website, perhaps.

Multimedia search as it stands is primitive.

1. Make it super easy for users to create and store media.
   
2. Make it super easy for users to find media.
   
3. Make it super easy for users to share and consume media, to display media on their own sites.
   

So the emphasis has to be on standardization of search so as to make it as easy as possible for search to happen. Maybe content providers and search engines should actively collaborate to that end. And the way to maximize sharing is to rely on the ad model. The average user wishes to be able to display for "free." And that is where the big money is, in the "free" zone.

Say if I took a photo, and uploaded it at some photo sharing site like Flickr, and I had the option to let just about anyone to display that image, as long as I could have that little link ad, that maybe links back to my own site, I would be okay. What about you?

Revver has a pretty good model for video.

Textbooks are game. Textbooks also have to be offered like media. Ad supported. Maybe disable copy and paste, at the author's discretion.

The average web user is going to want to produce content. Right now search is chaos. What is the GPS version of web search?

Advanced search options need to get much more intuitive.

Google is huge. It is an elephant. As it grows bigger, it is going to get less nimble. That will create huge markets for niche search. One size will not fit all.

Redefine search.

What am I looking for? How long does it take me to find it?



Some combination of machine search and social search might also work, especially when the end user is in control as to the degree of it.

It is to Google's advantage that it also provides platforms for content creation, like Blogger.

Just like it does with text, a good search engine should be able to help you search through multimedia without the multimedia sites actively collaborating. Multimedia sites should get indexed as they spring up, and they could spring up anywhere, any part of the globe.

Multimedia in English has been problem enough. The problem is compounded when you demand - as you should - that the multimedia search has to be language neutral. The machine should be able to understand any media in any language and offer it to you in your language of choice. That has to be the goal.

But first search has to get language neutral for text. Multimedia will be step two.

What if you are not necessarily looking for the most popular and most linked to and most visited sites to show up at the top. Some of the best sites for what you are looking for might be on page 350 or worse. How will you ever find them?

The advanced search option is not advanced enough, and it should not be a separate page where you basically have to fill out a form, it should be as intuitive as simple search. You tweak a few things here and there and you are ready to go.

Also how do you make it easy for people to do mashups and still not destroy the ad business model? How do you make RSS stuff really simple, as the name promises but has not delivered?

Create content. Consume content. The netizen demands those two basic functions.

And my pet peeve of all: why can't you "blog" in mathematical symbols? Like you can find chess players to play chess with at Yahoo Games, you should be able to collaborate on mathematical work. Just do a search and find them.

What does the netizen want? That will be a perennial question. It will still be asked a hundred years from now.

Remember that adage from the last dot com bust, that a site is all page hits and no revenue? Looks like page hits are all that matter after all. It was not a bad business model to start with. Just a little too early.

A Web 3.0 Manifesto
Dell, HP, Apple
Google Books: Primitive
The Next Search Engine
Memo To Bill Gates
Google And Languages
Internet Phones, Video Blogging, Nano
Google: Poised To Be The Number One Software Company In The World