A primer on key AI technologies
How long would it take you to write 500 personalized thank you cards after a wedding?
Could you write each one in the tone and style that you know would resonate with each guest based on your conversations with them, ranging from the bride’s weird Aunt Suzie all the way to Uncle Bill, that distant cousin twice-removed on the groom’s side?
Natural language processing enables marketers to generate on-brand, personalized messages at a scale never before seen, so the 500 thank you card project would take seconds, as opposed to several weekends.
“The benefit of using NLG-generated marketing messages and powerful statistical methods is that brands can deliver higher performing messages generated by NLG, but also understand how each component performs and which contributes the most to the outcome.”
— Panagiotis Angelopoulos, PhD, Persado Chief Data Scientist
Marketing leaders have similar challenges when it comes to delivering the right message at the right time to the right customer, and that’s where AI language technologies such as natural language processing (NLP), natural language understanding (NLU), and natural language generation (NLG) come into play.
Today’s CMOs and their staffs face a Herculean task—to aim creative messaging to large and diverse sets of consumers, but in a way that still creates a better customer experience, generates conversions, and increases product or service revenue. Only AI for marketing can help, which is partially why nearly 29% of CMOs’ 2018 marketing budgets went to marketing technology, according to the research firm Gartner. That number is expected to grow.
Learn more: The Power of Experimental Design to Deliver Marketing Insights
AI marketing technologies such as NLP and NLG help marketing teams personalize content and creative messaging at scale and at the rapid pace necessary in a digital world, yet retain the human feel consumers demand and the consistent brand voice marketers require. These approaches marry technical automation and sophistication with a human touch. Computer researchers and academics tend to view NLP as an umbrella field of AI that includes NLU and NLG. To better understand how they work and what they bring to marketing, however, it’s helpful to look at each separately.
Natural language processing technologies explained
Essentially, NLP algorithms are the text or voice input “readers,” while NLU and NLG are the “understanding” and “writing” components, respectively. Everything starts with NLP and then flows into increasing levels of the nuanced complexity that is human communication.
How to Use Natural Language Generation AI for Marketing
By Panagiotis Angelopoulos, PhD, Persado Chief Data Scientist
Persado created a unique and comprehensive database for marketing language that, far greater than just a dictionary, groups and tags words and phrases into hierarchical categories with words and phrases associated with each. For example, to express “Gratitude,” Persado collected thousands of ways to express it: “thanks for your interest,” “we appreciate your loyalty,” “our way of saying thank you,” etc.
These categories form the backbone of Persado-generated messages. By changing the positions, values, or words, Persado can easily produce a tremendous amount of language combinations.
Persado needed a statistical method to validate natural language generation’s efficiency, so it developed a proprietary machine learning method that uses experimental design with a small set of combinations and evaluates how each language category (emotion, position, format, length, etc.) and its values impact the outcome. Data scientists can not only view the performance of the overall NLG-generated message, but can also understand how each component in a message performs and which contributes the most to the outcome.
Persado has collected an enormous amount of marketing campaign data after working with clients from different industries for many years. And that enables the data science team to create predictive algorithms that produce high-performing machine-generated messages.
What is natural language processing?
Natural language processing is a convergence of computational linguistics and computer science. It is the primary method to analyze human language and break it down into smaller components, such as part-of-speech tagging, word segmentation, terminology extraction, speech recognition, or lexical semantics. Any middle school student who has diagrammed a sentence has witnessed firsthand how NLP algorithms approach their language analysis tasks.
What is natural language understanding?
Natural language understanding is the next step in the NLP process that leads to NLG. Once a machine analyzes human language at the “parts” level, the algorithm attempts to “understand” the communication’s meaning. That includes trying to interpret the sender’s emotion. NLU seeks to answer the question, “what is the intent of this message?”
What is natural language generation?
Natural language generation moves NLP from syntax analysis into the world of semantics: it produces meaning through content creation. The algorithms create words, phrases, and content based on processes similar to human methods, such as determining what information to include, logically ordering information presentation, using the appropriate grammar, and choosing the “right” verbiage.
Examples of natural language processing
NLP is the pervasive base that anyone and everyone who has to analyze text must use, including Siri, Alexa, text-to-speech capabilities (such as trying to send a text message using voice commands), Google Translate, email spam filtering, etc.: the list is almost endless. So, most consumers use NLP every day without realizing it when they interact with their devices’ speech-to-text translation functions. NLU is also used in any and all systems where humans interact with machines via human language protocols.
For example, Siri has to understand several aspects of what someone asks it to do. “Siri, where’s the nearest bank?” must be parsed so it understands what a “bank” is, comprehends that the command is a question, accesses geo-location, compares your location with the location of the closest bank, and then generates a response. As long as Siri doesn’t have problems deciphering word pronunciation (heavy accents, background noise, or speech impediments can present problems), then it will likely reply with an accurate response, thanks to natural language generation.
NLG continues to gain traction for use in generating long- and short-form content. In another example, The Associated Press has partnered with Automated Insights, the developer of the Wordsmith NLG platform. Prior to using Wordsmith, AP reporters generated only 300 stories on corporate earnings each quarter. After adopting the Wordsmith NLG, the AP now “writes” over 4,400 quarterly earnings stories.
Learn More: A Primer on Natural Language Processing and Natural Language Generation
Wordsmith doesn’t perform these functions completely on its own. AP and Automated Insights worked together to map editorial decisions and bake those into the Wordsmith algorithms. Over time, when machine learning if-then rules give way to more intelligent capabilities–such as generating its own editorial rules as AI detects shifts in reader responsiveness–the reports will become more personalized.
NLP enables much more than just email subject line optimization
One of the early use cases for NLP was to optimize email subject lines to improve direct response marketing campaign performance. While this is still an important use case and one where many brands find value, NLP has opened up never-before-seen possibilities for delivering value across the enterprise.
Business leaders in healthcare can now generate personalized messages that remind patients to take medication and get their flu shots because the AI system can process, understand, categorize, and generate the right message at the right time. Business leaders in operations can use AI to generate billing notices and collection notices that improve accounts receivable revenue.
The future of personalized marketing at scale depends on natural language generation. Otherwise, marketing, creative, and copywriting teams will spend their time writing the equivalent of millions of thank you cards in the race to be relevant and to achieve the proverbial one-to-one marketing.
NLG is a natural fit to deliver better copywriting and, contrary to the popular narrative that AI is poised to replace copywriters, human copywriters now use AI to augment and improve their creativity to deliver their best work.