Natural Language Processing (NLP) is a technology behind many of our daily interactions with the digital world. Every time you ask Siri for help with something or use Gmail’s autocomplete function to quickly compose an email, you’re using NLP in some form or another. At its heart, it is simply a form of AI that filters human language, extracts information from it, and makes decisions based on what it finds. Whether this is through text or speech, as part of an interaction or processing after the fact, any AI engagement with language falls under this umbrella. This could show up as an interactive interface like a chatbot, or simply as a way to make more accurate suggestions or spot trends. If you’re looking for ways to make use of this emerging technology in your own products and services, it is crucial to keep track of the latest developments in NLP to understand where these interfaces can be integrated into everyday engagements, particularly where there is an interface between a service and its customers.
The most visible use of NLP is in the customer service arena. While it is not a replacement for actual human interaction for the most part, what it can do is deal with all of the instances where a quick and simple query can be resolved using AI. The larger the customer base and the simpler the majority of queries, the more value it is able to add. Research firm Gartner, for one, reinforces this, predicting that by 2020, 85% of customer service interactions will be managed without any human intervention. The benefits of some form of 24/7 communication are also clearly felt by customers. According to Neil Patel, quoted in Forbes, 77% of customers have better perceptions of a business after chatting with them online, even if not to a human.
One of the subtler, but fundamental benefits of NLP, especially in the business space, lies in the way in which it enhances search. For much of our technological lives, we’ve had to search according the to criteria set in place by our systems. Use the wrong keywords and you get the wrong results. NLP has changed this by adding a layer of interpretation to the process, searching for what it thinks you meant, not just what you said. Klevu, for example, is a popular smart search offering that can be integrated into e-commerce sites. It uses NLP and machine learning to allow customers to search for products using everyday language. So instead of customers having to use strict criteria to find products, Klevu draws on synonyms and context cues to work out what people mean and then delivers more relevant results. This leads to more purchases for a start, but also allows for brands to gather data around what people are searching for and how their assortment could better serve the customer needs.
There are naturally going to be reservations when it comes to the implementation of NLP in customer-facing offerings. Data security is always a key fear, with recent concerns raised around how seriously both Amazon’s Alexa and the Google Assistant take customer privacy. Resolving issues around security, however, is a relatively simple step when implementing NLP, and comes down to general good data security practice.
The more complex issue to deal with is the alienation people can feel when engaging with NLP – a kind of auditory uncanny valley. One underlying fear is that it only further serves to provide another way to avoid connection with a real person. What’s more though - and a phenomenon which became immediately apparent after Google first demoed its new Duplex feature at the 2018 I/O keynote - is that people have an extremely antagonistic reaction when they feel they are somehow being “duped” by the machines into thinking they’re conversing with a real person.
Personally, I don’t see this as something inherently problematic, however. It is less a function of how the system itself is built and more the societal nuance that will need to be observed if this sort of technology is ever to achieve mainstream adoption. I see particular value in creating hybridised models that use NLP and human engagement to best interact with people. For example, when a financial services company with a large volume of inbound calls worked with us recently, we developed a system to effectively screen queries using NLP to ascertain the best way to direct them to a real person. This kind of “robo-advisor” model doesn’t replace the actual call centre agent, but allows for more specialised staff to be retained, using NLP to direct customers to the right place upfront and thereby minimizing waste and improving customer satisfaction through reduced call wait times.
Ultimately the two true benefits of using NLP lie in helping the user get to where they need to go more quickly and with less effort than before, and ensuring that as a business you are able to learn from this journey to develop a richer experience over time. In a linguistically diverse country like South Africa, technology such as this can be incredibly empowering for both brands and customers. It enables more natural channels of communication to be opened up, without the need to hire extensive resources to fulfil the task. In the end, what it comes down to is the simple fact that if you need something done it will always be far more efficient to just ask.