With Voice Search, users can speak search queries into their mobile phones in English, Afrikaans and Zulu, and have their queries relayed back as text, and increasingly, as speech.
No ad to show here.
“Voice has always been the most natural way to interact with a phone, since speaking is typically faster than typing,” said Google’s communication manager for Africa, Julie Taylor.
Speaking at G South Africa in Cape Town on Monday morning, Johan Schalkwyk took the stage to demonstrate the power of Voice Search. Having just flown in from London, Schalkwyk demonstrated how he checked his flight details by speaking the airline and flight number into the phone, which returned the departure and arrival times of the plane.
Schalkwyk is probably the world’s foremost voice recognition experts. He left South Africa in the early 90’s for Oregon, USA to begin his specialist work in speech recognition techniques, and joined Google in 2005.
Once he had pinpointed his flight details, Schalkwk asked Google to check the “weather in Cape Town” and it quickly returned the results successfully. But he noted that the results were in Faherenheit, so he “asked his phone” to convert 63 degrees Fahrenheit to Celsius, and it obliged to claps and cheers from the audience. Admittedly, performance was a bit slow due to the challenged quality of the wifi at the event, but otherwise the functionality performed perfectly.
Voice Search is likely to be well received in the marketplace. Wikipedia summarises the history of voice search like this: “Google Voice Search was a tool from Google Labs that allowed someone to use their phone to make a Google query. After the user called (650) 623-6706, the number of Google Voice’s search system, they would wait for the words Say your Search Keywords and then say the keywords. Next, they would either wait to have the page updated, or click on a link to bring up the search page the user requested.”
But the technology has changed quickly. Schalkwyk explains that “the Voice Search app works by digitising voice in real time, then transmitting it wirelessly to Google’s servers. The data is analysed by speech recognisers using various acoustic and pronunciation models which analyse the words that you say and returns the search results as text.”
One of the key challenges for voice search in an African context is the lack of available content. So while Voice Search may understand the query perfectly it is not always able to deliver applicable results, because many times, those results do not exist.
But the timing seems perfect for South Africa, as Voice Search is being launched on the same weekend that South Africa’s largest-selling newspaper, the Sunday Times, launched a Zulu-language edition that should dramatically increase the amount of Zulu language content on the web.
The technology is expected to be used when people are out and about, in restaurants, walking or in the car, so it needs to be customised for different acoustic environments. Professor Etienne Barnard from the University of the North West was in charge of collecting voices, accents and vocabulary from across South Africa to populate the database.
The technology doesn’t always work perfectly. Numerous attempts to look up Professor Barnard’s profile returned results for “MTN Final” and “Themba Modi”, amongst others, but the technology points the way to a world of increased functionality for the visually impaired and the functionally illiterate who cannot rely on the written word for their search results.
Schalkwyk notes that more South African languages are being incorporated into Google Voice Search right now, and sees connectivity as being the biggest challenge facing the adoption of this technology.
