Speech analytics can be divided into two categories: speech recognition that is able to recognise and understand spoken words and phrases, and speech analytics that is able to recognise the emotion behind a speaker's words. Both of these types of speech analytics are used to gather information from verbal interactions with customers that may not be captured otherwise.
The speech analytics process is made up of many steps. The first step is to convert captured audio into data. This can be done using three methods: phonetic approach, large-vocabulary continuous speech recognition (LVCSR) and direct phrase recognition.
The phonetic approach uses phonemes (a distinct sound that makes up part of a spoken language) to identify words. This method is fast when it comes to processing, but can result in false positives for words or phrases that sound similar and is slower.
The LVCSR method is based off a phonetic system, which then feeds into a dictionary to create a transcript. This method uses statistical analysis to ensure that a sentence is correct, based on past use of that combination of words. The first issue with this method is that processing the audio takes longer since it has to be run through a vocabulary. Another difficulty of this model is that words need to be contained in the dictionary in order to be detected.
The last method is direct phrase recognition, which understands pre-defined phrases. This method is the most reliable due to the fact that that there is no conversion, which might result in data loss and the length of the phrases which makes them easier to identify.
The evolution of speech analytics
Many of us no doubt remember the first time we spoke into a microphone attached to a computer and it responded with a very robotic voice. Some of you may have only experienced voice recognition and speech analytics with the likes of Siri when you got your first iPhone.
However, the roots of speech analytics go all the way back to 1952 with the Audrey System.
In the report
From AUDREY to Siri
, Robert Pieraccini discusses how the Audrey System was built in 1952 by Davis, Biddulph, and Balashek at Bell Laboratories. This device was able to recognise strings of digits with pauses in-between with amazing accuracy if it had been configured to the speaker’s voice.
Pieraccini continues to discuss the advancements made with voice recognition and speech analytics. He highlights how Siri was a massive breakthrough for multiple reasons, such as:
- Its almost infinite vocabulary
- The ability to understand the context of language
- It was perceived as intelligent
- Interacting with it was fun
This has led to its use in many areas, such as:
- Providing customers with upfront answers, such as a train schedule or how much a specific service costs
- Routing simple calls - when a customer has a specific query, putting them through to the right department
- Automating customer experience with bots - AI and speech analytics are increasingly used in devices like Amazon Echo to power virtual assistant Alexa or contact centres to augment the services of contact centre agents
Pieraccini, R. 2012 From Audrey to Siri, Pg 6-7
Uses of modern speech analytics in the contact centre
This evolution in speech analytics technology in the contact centre has led to the use of speech analytics as a way to improve customer experience in different ways.
The first way this is used to improve customer experience is by capturing historical call information - that is recording, transcribing, and analysis of customer calls. Being able to capture this information and analyse why and how your customers are engaging with your contact centre agents are vital in improving customer journeys, and therefore improving your overall customer experience. This analysis can also be done to identify ways to cross-sell or upsell to your customers.
Another way that speech analytics can assist contact centre agents is by doing live analytics of the pace, stress, tone or a number of other vocal elements in order to improve a customer interaction by either predicting the nature of a customer’s call (for example, if they’re calling about a complaint) or by guiding the contact centre agent (for example, to slow down or to clarify if the customer sounds confused).
Regardless of how you implement speech analytics, the overall goal is to gather information to improve the quality of service for your customers and to help your business grow.Merchants
is a customer experience provider who can assist you with speech analytics and digital transformation. For more information on digital transformation, contact