What is the language used to enable talking to Samsung Galaxy’s speech recognition?

Heidi Johnson

We've all had a go at trying to talk to a phone that recognises speech in one format or another, but when it comes down to it, trying to get a mobile devise to understand 'please call Rachel Johannesburg' is easier said than done.

Nuance Dragon is the company who make the majority of speech recognition apps and software. They describe it as in Nuance Voice Control 2.0 'the use of speech to access and utilise mobile handset-based or network-dependent features, applications, services and content.' They do not classify what kind of specific speech to use, so how can anyone know?

How does this type of artificial intelligence work? In order for a mobile to be able to understand the human speech the computer is essentially spoken into. The creator of the device or software talks into a microphone connected to the computer, and then translated for the computer to understand. Whoever speaks into the computer determines the level of accuracy in the language the computer responds to; this tends to be smart intellectuals with a formal British accent. The phone is also set to Oxford English (available to change), so when a slight accent is spoken to the speech recognition the mobile doesn't recognise the change of pronunciation.

As Victor Keegan wrote in The Guardian, Has voice recognition finally come of age? In 2007 'the faster I spoke the less effective' I tried out for myself a few different sentence structures to see what was recognised. I asked 'May I call Hannah?' using a modal verb to see if the speech recognition would understand this form. The response I received was an app for an email to be sent. After trying a few more commands with hedging in, it was evident Samsung didn't understand me.  So instead I looked at the examples given by the phone, they only consisted of commands. A large majority of the examples showed command in theme position; however, 'Facebook Update' and 'Note to self' were in my mind not clear commands and look like a sort of outlier.

Norman Fairclough has said in Language & Power 2^nd edition 2001 that there is likely to be a 'general requirement for consistency of language forms' when talking formally. As I'm sure most are aware, when talking to any form of speech recognition the word needs to be clear and precise, pronunciation is key, avoiding slang, taboo and unusual sentence forms.  Fairclough goes on to explain that the 'vocabulary used in formality must be selected from a restricted set throughout' similarly to this Samsung explain in The Next Communication Experience  'you can control the Samsung Galaxy SII with a few select words.' This gives a strong indication that speech recognition is formed on formality.

If the speech recognition software is set to a formal formatting then it would explain why unusual vocabulary and slang are not understood and converted into what is considered to be formal Oxford English.

References:

http://www.nuance.co.uk/for-business/by-product/nuance-voice-control/index.htm

www.samsung.com/global/microsite/galaxy2/html/feature.html

www.guardian.co.uk/technology/2007/dec/13/research.comment?INTCMP=SRCH