It’s an evolving field of study and development predicated on vocal user interfaces. It aims to accelerate translation, eliminate frustrating customer experiences and enable people to speak directly to computers and receive responses in a more natural, human way.
What is it?
Computational linguistics is a multidisciplinary field that seeks to digitally model natural language — helping computers find new ways of understanding the spoken word, and generating their own ‘spoken’ responses.
It has important implications for both customer and employee experiences. Computational linguistics wants to iron out the recognition and interpretation limitations of basic voice recognition technology, and enable all of us to communicate with computers in a friction-free way that feels as natural as talking to another human.
What’s in for you?
The biggest advantage of computational linguistics is that it’s helping to improve the quality of automated customer and employee experiences. Tools like chatbots for example are great at driving down customer service costs, but until now, have been limited in their ability to respond naturally.
Computational linguistics helps solve that problem, through pushing the envelope on automated experiences, moving away from the notion that these should just be delivered at a low cost. Instead, they’re intended to be engaging and don’t drive customers away in frustration.
But its capabilities go beyond just removing the frustration from automated customer interactions. It gives machines a deeper understanding of customer intent and sentiment — helping you direct them to the right place faster, and provide them with better-targeted recommendations that drive revenue while increasing customer satisfaction.
What are the trade offs?
Understanding and modelling human language is complex. It takes a lot of data, and a lot of processing power, which makes it prohibitively expensive and complex for most businesses. Cloud costs for training a single model using the most popular open data sets can easily extend into the millions. As a result, there are currently few viable use cases.
In the past, voice recognition technologies have been found to show significant bias in their ability to understand people and dialects across diverse cultures. While computational linguistics itself seeks to solve that problem and remove the bias, trained models are only as good as the data they’re given — making it easy for bias to creep into even the most advanced use cases.