If you want to learn a new language, the options are not lacking, and they are free. But one aspect of the process that has been overlooked is pronunciation, which is a particularly important part for professionals. Online, pronunciation learning usually consists of "listening to a recording and then repeating it (in an empty room)." A new platform, Blue Canoe Learning, uses an established program and machine learning to make easier and more effective things. This is the first company to join the new AI2 incubator in Seattle, and raised $ 1.4 million to expand its operations.
There are millions of learners in English, many of whom speak it as their first language, but who are still unable to make themselves understood especially to Americans, who are particularly under-exposed to certain accents.
The reality is that American English is the international language of many industries, and no matter who in any of them, that it 's safe. Acting call center employees, software engineers or executives, can benefit from a US focus. . Their employers also know, Blue Canoe goes straight to them instead of taking a direct approach to the consumer.
I say potato …
The problem with teaching pronunciation is that it's not just about telling people how to say something correctly, but to hear how they say so and provide corrective advice. This kind of personal comment is difficult to scale.
This is particularly difficult when one considers the neurological limitations of adult learners. Unless you learn some sounds at a young age, your brain ends up getting rid of the mechanism to hear them, which makes it difficult for some speakers to understand that they are saying something incorrectly. (I've recently had this experience in China when I tried to get instructions for Futian, I understood it correctly at the 8th or 9th test.)
Add to that that in American English, you have 5 vowels but 14 vowels, and that's a recipe for confusion.
"People have not focused on the last mile," said Blue Canoe founder Sarah Daniels. "They say they're talking, but it's glorified, listen and repeat."
An attempt to solve this problem, offline anyway, was the color vowel system, which relies on mnemonics and rhythm to help unlock sounds in your brain that you do not even know are there . Each vowel is associated with a color and alliterative sentence: green tea or brown cow, for example. So, to learn how to say "speed," learners would be asked to say "green tea speed." The combination of repetition and color, in theory, helps with retention and helps to produce the vowels in question.
I say theoretically not because I personally doubt it, but because there is not much literature on it; pronunciation is a difficult thing to measure in relation to vocabulary or written competence – very subjective. But the Peace Corps, the State Department and several major universities have adopted the system, so until studies come out, I agree with their judgment.
Scaling up the system
Blue Canoe (itself a mnemonic phrase) worked to scan the color vowel system and pack it as an application. It is still at a very early stage, with more content expected as the company learns from its pilot programs.
Users play a deck of cards (the first of several games and activities to be included) that forces them to say the word of vocabulary on the card they play; an automatic learning system listens and identifies if they have pronounced it correctly and, if not, provides relevant comments.
At first, I thought that the system would have been driven on reams of Anglo-Saxon data in English, and analyzed the delta between waveforms, but it's smarter than it. Instead, Blue Canoe had people with different accents speaking, and their pronunciation was annotated word by word by professionals. So an "r" pronounced with a roll (by, for example, a French speaker) would be treated differently from an "r" pronounced closer to "l" (by a Japanese speaker). Highlighting a different syllable from Americans (the most common difference) will also be detected.
The number and type of errors also allow the application to create a global note and highlight the words or sounds on which the speaker improves, needs help etc. Part of the plan is to track these assessments and compare them to in-person professional assessments to validate them as an automated and objective score of a user's pronunciation progress. That alone would be a useful tool for businesses, but the opportunity to improve that score, of course, is also appealing.
Blue Canoe already works with several companies to create special programs with vocabulary and objectives (and perhaps activities) tailored to their needs – technical terms, practical phrases, etc. This pilot program is expected to last a few months, then next quarter should see a more public deployment, perhaps with the documentation of the evaluation process and the effectiveness of the application.
Kernel Labs led the $ 1.4 million round, but Blue Canoe will follow its upcoming months of guidance and support from the Allen Institute for AI, which has established more early this year a new low profile incubation program. This is their first selection for a company to adopt and invest, although it is not necessarily representative of the type they are looking for: I had envisioned it, when I was spoke with AI2 earlier, a kind of wild-haired geniuses who needed established AI brains. But it works too.
Stock Image: Bryce Durbin / TechCrunch