Speech analytics: are you listening?

Speech analytics delivers priceless customer information. The technology has been around for some time, yet is now gaining greater status in the contact centre due to smarter apps being developed in response to other current and major IT trends, in particular, big data.

Ebrahim Dinat, chief operations officer at Ocular Technologies, a South African contact centre solution provider, says speech analytics offers businesses a critical customer service tool, and emphasises that contact centre managers should take heed of the strength this type of technology can play in the entire business.

"Today, new and innovative speech analytic applications have come to the fore. This software has the ability to change the scope of the contact centre environment and noticeably improve customer service. With massive amounts of data having created an intimidating whirlpool of information, speech analytics can help to forge a clear path through the swirling data, grab information and make sense of it, and as a result, establish a more efficient customer centric business," he says.

Speech analytics is generally defined as the process of analysing recorded calls to gather information, bring structure to customer interactions and expose information buried in customer contact centre interactions with an enterprise.

DMG Consulting, a leading independent research, advisory and consulting firm, puts speech analytics as a "true change agent", and in its synopsis of its "2014-2015 Speech Analytics Market Report and Consulting Services", says "the contact centre speech analytics market, in its 11th year, delivered exceptional performance in 2013".

The abstract also states that DMG has long praised these solutions for their ability to help organisations realise significant tangible business benefits. "As speech analytics solutions continue to evolve, users realise new and greater returns. This success will only increase. The number of reported contact centre speech analytics seats grew by 26%, from 2 292 733 in July 2013 to 2 889 031 as of the end of May 2014. The vast majority of these seats are from sales of post-call solutions, although the adoption of real-time speech analytics is beginning to pick up," it says.

Adding fuel to the speech analytics popularity fire is the world's largest market research store, Research and Markets, which announced the addition of the "Speech Analytics Market by Solutions & Applications - Worldwide Market Forecast (2014-2019)" to its offering.

Its media announcement states, "...the speech analytics market is expected to grow from $456 million in 2014 to $1.33 billion by 2019, at a compound annual growth rate (CAGR) of 23.9% from 2014 to 2019. LA and MEA regions are growing rapidly with CAGRs of 38.2% and 35.6%, respectively, from 2014 to 2019."

"The speech analytics market is growing rapidly and fuelled by various trends, including the emergence of call centres, increasing focus on regulatory requirements, and market intelligence and competitive intelligence. The real-time analytics is the important product trend that fulfils the regulatory requirements in the government and public utilities and other verticals. These verticals are expected to have an optimum market share in the overall market."

Dinat shares a report by its software partner, Aspect, called: "Speech Analytics - It's all About Statistics". In it, the company takes the stance that when considering implementing a speech analytics solution, there is a fundamental decision to be made by organisations before all else: "Which of the two major approaches to speech analytics should you choose... phonetics or LVCSR (large vocabulary continuous speech recognition)?"

Written by Aspect's speech analytics expert, Bob Moore, he explains that both phonetics and LVCSR approach process incoming speech by identifying phonemes (unique elemental sequences of sounds), and then identifying words, phrases and even sentences that were spoken. "This is, by definition, a statistical matching process, which means that sometimes the matches are wrong, introducing errors into the remainder of the process, whether phonetic or LVCSR. The differences between the two approaches are all about how they reduce or eliminate those matching errors," he says.

Moore goes on to highlight that although both approaches will retrieve recordings that do not contain the desired word/phrase, and both approaches will miss recordings that do contain the desired word/phrase, the critical difference between the two is: In the case of the LVSCR solution, the audio has already been translated to text at the time of your search.

"When you search on 'wireless', you get only those audio clips that these error-prone approaches identified as 'wireless'. You might actually be getting clips of someone saying 'worthless' or 'we're less', and you will be missing clips from people with unusual accents. In the case of the phonetic search, the system has not translated the phonemes to something as rigid as a textual word that is either matched or not. The phonemes continue to exist as part of the search process, so you can set a 'confidence threshold' at the time of your search. You can set it to low confidence if you want to get all of the occurrences of a word/phrase, or you can set it to high confidence if you want to get only the occurrences of a word/phrase," says Moore, who at the end of the report believes that LVCSR is a better solution for applications like transcribing a television show for closed captioning.

"However, if you have a large number of speakers with different accents, poor audio quality and conversational unscripted speech, and you need to find events, phonetics is a superior solution," he says.

Dinat adds that Moore's case for choosing the right speech analytics application might seem like a very pedantic matter to organisations, but it needs to be kept in mind that speech analytics can be used for various purposes in a business, such as compliance or a representative sample of a product/service checks.

"The question every company cannot afford not to ask is: are you listening to the priceless information your customers provide? This information is power! As Virginia Rometty, IBM CEO, so aptly stated: 'Data will be the basis of competitive advantage for any organisation'," concludes Dinat.