Automatic speech recognition technology is slowly taking over many mundane call center operations
THE next time you hear about the slowdown costing hundreds of call centre jobs around the world, at least a part of it could be the handiwork of Automatic Speech Recognition (ASR), a simple but effective technology that uses computers to identify and analyse spoken words. The technology has been around for a while — even your home computer has it –and is what sets a ringtone when you say the first line of a song or tells you the status of your train as soon as you speak its number into a phone.
In India, the technology has been used extensively over the past couple of years in the telecom, healthcare and automotive sectors. But things are a bit more complex here considering the plethora of regional languages in vogue. But Lattice Bridge Infotech Private Limited (LBIT), the Chennai-based company which pioneered ASR technology in India, has made this very challenge its USP by providing voice-based solutions in 11 Indian languages along with Arabic.
LBIT managing director C Mohan Ram says the ASR and related technologies were initially designed only for English, with regional language support almost non-existent. LBIT, however, changed this with its Simply Speak application for 11 Indian languages. “India has 300 million mobile phone users compared to just 4 million broadband connection, this shows that the spoken word is still supreme here,” says Ram.
He says ASR tries to bridge the communication gap, more prominent in India due to the rampant illiteracy in many parts, by providing services to the common man irrespective of location, time and language. “The application pushes all complexities to the back-end, keeping the front-end interactions simple,” he explains. Simply Speak can now handle live conversations in Hindi, Tamil, Kannada, Telugu, Malayalam, Gujarathi, Marathi, Punjabi, Oriya, Bengali as well as the Indianised English.
Though his company has clients in the BPO industry in the Middle East as well as South Africa, Ram is quick to clarify that the technology is still not good enough to replace call centres. “It can only help make some operations cost-effective. More complex tasks will still need the brain power of people. However, we can make their lives less boring by taking over mundane operations,” adds Ram, whose company has been recognised as among the top 10 centres of excellence in ASR technology across the world. But there is not denying the fact that ASR is cost effective as it can get to work without any requirement for personnel training, especially when faced with human resource or skill shortage.
Ram says his company has achieved accuracy levels of up to 95 per cent for English and between 79 and 92 per cent for the regional languages. The accuracy is measured as FARR (First Attempt Recognition Rate) for various languages and specific application.
However, the local dialects have only complicated things. “Hence, localisation and regionalisation is a continous activity. We spend a lot of time collecting corpus and tuning of various languages,” he adds.
Ram believes the ASR technology has seen a lot of development over the last decade. But work is on to build a larger corpus of words and understand the nuances of localisation and regionalisation.