Before OpenAI’s GPT-3 burst onto the scene a few years ago with the first commercially available large language model (LLM), the modern consumer’s most likely interaction with the category of AI known as natural language processing (NLP) was customer service agents like “Ask Julie.” Amtrak introduced the virtual travel assistant in 2001 as an early experiment in NLP. Her cheerful demeanor has made Julie into a success story in automated customer service delivered with a human touch. Amtrak earns 30% more revenue on 25% more bookings than it could without Julie’s technology handling the bulk of the call volume.
As the umbrella field of AI dealing with language, NLP has come a long way since the early years of Ask Julie. A major leap in its capabilities came with the emergence of large language models from organizations like Anthropic (Claude), Google (Gemini, formerly Bard), OpenAI (GPT-4 and its associated applications, such as ChatGPT, DALL-E, etc.), and others.
The generative pre-trained transformer method for training natural language models accelerated the speed and scale with which companies could develop them. As a result, there are now more LLMs and a vast array of new companies emerging that take advantage of NLP to solve business problems.
With this post, we dig into the facts about NLP to review how it enables the most powerful Generative AI content applications.
NLP technology converts human language into structured data that a computer can interpret. Think of NLP as allowing a computer—which technically understands only ones and zeros—to read text written in normal human language.
Before NLP, all the queries that came into an organization, from emails to call center requests, had to be read or listened to by a human agent because only humans could see that the words were words. Just as challenging was the fact that the content of all those interactions couldn’t be easily captured and analyzed to drive improvements.
NLP changes that. By providing a way to capture language and convert it into interpretable data, NLP vastly improves the speed with which companies can take in information or requests as humans usually deliver them: in natural speech.
The field of NLP has existed for decades—it is not new. Yet data scientists are exploring new techniques and innovations all the time. In the 2010s, a group of Google researchers discovered a technique that led to “Generative Pre-trained Transformers” (or GPTs). The current generation of large language models underpinning Generative AI are GPTs.
GPT-based language models have an architecture that lets them evaluate words both individually and as part of a full sequence, as in a sentence, or even a paragraph. This allows transformer-based language models to ingest and process larger volumes of training data at once. The process is much faster than it would be if they had to interpret each word individually, both for itself and its placement. Transformer model training thereby produces much larger and more syntactically accurate models than previous NLP methods. Transformer models are also self-learning, which means they do not require the same volume of manual tagging and scoring from human workers as NLP models did in the past.
Most of the NLP solutions that deliver business value include complementary functionality such as natural language understanding and natural language generation.
Whereas NLP is the functionality that ingests language data and makes it interpretable by a machine, NLU is the functionality responsible for identifying what the speaker or writer means.
Applications need this interpretation layer because language is all about context. Formal language by itself has its nuances and points of confusion. Add in slang, differences in pronunciation, and regional accents, and the difficulties multiply. NLU’s job is to parse these nuances to understand what the speaker or writer is saying or asking.
Natural language generation joins NLP and natural language understanding to deliver the output part of the trio. Specifically, NLG is the piece of the technology responsible for producing verbal or written text that sounds like a human would speak or say it.
When integrated with the Interactive Voice Response (IVR) software used in call centers, with personal assistants like Siri or Alexa, or with modern Generative AI applications ChatGPT and even Persado, NLP technology takes in content in the form that consumers normally speak or write using natural language. The system does the work to interpret their words (NLU) and route them to the right human agent, turn on the requested function, or generate a response using NLG capabilities.
The lines between NLP and its cousins are invisible to an end-user of a Generative AI solution like ChatGPT. Or like Persado Essential Motivation, an enterprise Generative AI trained for marketing teams to craft copy designed to motivate customer action.
When you write a prompt into any Generative AI application, you don’t need to think about whether the solution is processing, understanding, or generating language. All you care about is fulfilling your task. Behind the scenes, however, all three are necessary to turn a prompt into a fluent and effective output.
In the case of Julie, she listens as a customer explains what they need, interprets the request, finds the answer, and then delivers her response back in conversational language. Need to book a ticket to Boston? “Sure, I can take care of that” is NLP in action.
Customer service tools like Ask Julie were among the earliest applications of NLP. In business, NLP allows organizations to automate customer interactions through call centers, website chatbots, and email, and still maintain a human touch.
Since the launch of Generative AI, however, the use cases are expanding rapidly. Chatbot marketing has received a significant upgrade, for one. There is now a vast array of potential applications of the technology. Research by McKinsey finds that 40% of enterprise leaders are integrating Generative AI into their workflows for at least some tasks. Based on estimates of continued rapid adoption, Goldman Sachs estimates that Generative AI will drive a 7% increase in global GDP and boost global productivity by 1.5% over the next ten years.
The use cases span well beyond customer-facing activities. Think of team members in the finance department leveraging NLP via a custom Generative AI to produce reports. They can efficiently create different versions for regulators and for the board of directors using the same sources of data but each with a different emphasis.
Another use case is retailers writing product descriptions geared to different audience segments. Or sending personalized email newsletters with different headlines and a different mix of content to appeal to the interests of the customer group.
Marketing leaders are actively investing in NLP through their embrace of Generative AI. Among other strategic priorities, these investments are boosting marketing’s ability to personalize content and create messages at the scale business requires today.
From customer service to customer engagement, NLP via Generative AI is providing brands with the tools they need to communicate with more customers in ways that marry technical automation with a human touch. Just ask Julie.
And to explore your options for a risk-free trial of Persado Motivation AI, ask us about starting a risk-free trial.
The post AI 101: Natural Language Processing (NLP) appeared first on Persado.
Understanding what's happening behind large language models (LLMs) is essential in today's machine learning landscape.
AI accelerationists have won as a consequence of the election, potentially sidelining those advocating for…
L'Oréal's first professional hair dryer combines infrared light, wind, and heat to drastically reduce your…
TL;DR A conversation with 4o about the potential demise of companies like Anthropic. As artificial…
Whether a company begins with a proof-of-concept or live deployment, they should start small, test…
Digital tools are not always superior. Here are some WIRED-tested agendas and notebooks to keep…