Subscribe to Newsletter

Thank you.

You are now added to our subscriber list.
Oops! Something went wrong while submitting the form.
Please check your details and try again.

De-mystifying NLP: What is it, and how can businesses get to grips with it?

By Soffos Team
April 25, 2022

At Soffos, we believe that language is at the heart of human intelligence.  

This is an idea that has been long been passed around in academic circles, and one that has made its way into the field of Artificial Intelligence (AI) in the form of natural language processing (NLP) over the years. This refers to a branch of AI concerned with giving computers the ability to understand text and spoken words in much the same way that human beings can. It is hoped that one day, when these technologies reach full maturity, computer systems will be able to understand and converse with individuals in a manner indistinguishable from how humans interact. However, at present, we are quite far away from achieving this milestone. In fact, until recently, AI has had a language problem.

Obviously, human language is filled with ambiguities, metaphors, differences in tone and intentionality, in addition to homophones and homonyms that complicate matters further. With this in mind, the task of building machines that understand the unique processes of human speech and semantics has proven elusive, such that experts have deemed it an ‘AI complete’ problem. To call a problem ‘AI-complete’ means that it cannot be solved by a simple specific algorithm – in other words, to solve language is to solve AI itself.  

Thankfully, academic researchers are making some real progress in solving this ‘AI-complete’ conundrum. Today’s AI can now correctly answer medical queries and pen articles reflecting on its own limitations. OpenAI’s GPT-3 can even crack jokes and craft its own poetry!

So, what breakthroughs have been made, and how can businesses plan ahead for real-life adoption?

Back to basics: what is NLP?  

Essentially, NLP combines computational linguistics – rule-based modelling of human language – with statistical, machine learning and deep learning models. In practice this means that these technologies can enable computers to process human language, whether this is the form of written text or voice data, to ‘understand’ its full meaning, complete with the intentionality, sentiment and context of the person speaking or writing. Moreover, NLP drives computer programs that can translate text from one language to another, respond to spoken commands, as well as rapidly summarize large portions of text.  

If all of this sounds unfamiliar, chances are that you have already interacted with NLP: if you have used Gmail, you are likely to be familiar with the autocomplete function that predicts what users will write next to speed up the email-writing process. Likewise, customer service chatbot, GPS systems, digital assistants and speech-to-text dictation software are all underpinned by NLP technologies – so perhaps it isn’t remote as many may think!

Recent breakthroughs and solving the language conundrum

Increasingly, NLP is playing a growing role in the creation of enterprise solutions to help businesses streamline their operations, while boosting employee productivity. Certainly, this is a niche that we are aiming to carve out at Soffos, as we work on honing our algorithms. In the industry more generally, two related technology breakthroughs are making waves at the moment: self-supervised learning, and a formidable new deep learning architecture known as the transformer.

The transformer first came to prominence back in 2017 in a paper declaring that “Attention Is All You Need.” Prior to the introduction of transformers, most NLP technologies were based on recurrent neural networks, which process data sequentially. That is, one word at a time, in the precise order that the words appear. In contrast, the beauty of transformers is that they make language processing ‘parallelized’. This means that all words in a body of text are analyzed at the same time, thanks to an AI mechanism known as ‘attention’. Adding to this development, self-supervised models mean that training on far larger, unlabelled datasets is now a possibility.

What’s next for NLP?

Already, NLP is the driving force behind many different real-world applications. We’ve mentioned a few in this blog so far, but individuals can expect the likes of chatbots to become more sophisticated over the years. Likewise, companies will no doubt be able to glean valuable marketing insights from social media sentiment analysis tools that extract the attitudes and emotions buried that are in text, in response to products and emotions.

In the realm of language translation, machine translation technologies are making magnificent strides forward when it comes to accurately capturing the meaning and tone of user input. However, perhaps the most exciting avenue for NLP is its capacity to transform the EdTech sector. From our perspective, this could help businesses to consistently train their staff on the go, in a way that takes into account individual ways of speaking, as well as interpreting improving instructional methods for teachers in the classroom.

If you’re interested in learning more about the brains behind the products we have in the works, head over to our technology page where we explain the nitty gritty of our plans to democratize learning!