3 Things You Didn’t Know About Speech Analytics
Contact centre speech analytics is not a new concept
Around since 2002, recent figures show this technology is currently in use in roughly a quarter of organisations.
The sectors that account for some of the highest rates of speech analytics implementation include insurance (54%), medical (45%) and retail (40%) – showing that it’s a popular product that must produce results.
However, there’s still a considerable amount of information that companies do not understand about the technology – studies show up to 43% of companies do not yet know what speech analytics really is or how it can benefit their business. We’ve highlighted some of the common misconceptions, in our latest list of 3 Things You Didn’t Know About Speech Analytics.
1. The ‘traditional’ contact centre is no more.
Contact centres are booming. The stats speak for themselves; the number of telecommuters has increased by 800% over the past five years. With more and more agents working remotely, the ability to effectively and efficiently manage them becomes more difficult.
This is where speech analytics comes in. This technology is capable of turning unstructured audio data into actionable intelligence enabling management and sales leaders to monitor and analyse staff performance, no matter where they’re calling from.
2. Speech analytics in numbers.
In 2015, a report from Contact Babel entitled “UK Contact Centres in 2015: The State of the Industry & Technology Penetration Rates” report shows there are 5,840 contact centres in the UK, with 734,000 agent positions.
It also shows that in 2014 there were over 42 billion minutes of inbound calling in the UK alone. Lastly, this report states that 1 in 8 UK contact centres were looking to implement speech analytics in the next 12 months; a figure set to rise as the technology develops.
3. The various types of speech analytics.
It’s forgivable for a novice to confuse this type of technology, it’s complex and works in many ways. While there are similarities between each, they are in fact very different – and there are three main approaches to speech analytics and how it works.
This is the fastest approach for speech processing, using phonemes as the basic unit of recognition. It’s fast because the size of the grammar is very small. In most languages there are only a few tens of unique phonemes, making the output of this recognition a stream (text) of phonemes, which can then be quickly scanned.
LVCSR (large-vocabulary continuous speech recognition)
This system processes speech much more slowly, as the basic recognition unit is a set of words (bi-grams, tri-grams etc.). This means it must have hundreds of thousands of words to match the audio against. However, the speech is much richer to work with as the output is a stream of words.
This enables the software to identify new business issues, raise queries are much faster, and the accuracy is higher than the phonetic approach. As the complete semantic context is in the index of the software, it’s possible to find and focus on call issues very rapidly, helping managers to improve performance.
Direct Phrase Recognition
Rather than first converting speech into phonemes or text, this approach directly analyses speech.
A set of specific phrases is pre-defined as being worthy of identification, and the speech analytics software will recognise these phrases. By using this approach, no data is lost in conversion – therefore this method generally provides the highest data reliability.
If you require any more information on speech analytics software, or would like to discuss your contact centre’s needs with a specialist, call Daviker today on 0843 634 5821.