Teaching AI the Language of Your Business

Most of us interact with some form of AI on a daily basis, both at home and at work. Every time we do a Google search, use Siri, book a ticket, get a Netflix recommendation, or receive a credit card fraud alert, we are interacting with AI-enabled applications. It is also often an AI-enabled function that decides to approve or deny a social service or a credit application, to reject a job application, and to file an email as spam. With faster computers and better algorithms, more and more tasks are being automated, increasing productivity and efficiency, and one area that has seen great advancements in the past couple of decades is natural language processing (NLP). (See “Basic Concepts” for a definition and examples.)

NLP helps machines understand human language. The ability to process natural language has given rise to voice-based assistants, machine translation services, analysis of social media for product management, intelligent content management and extraction, and many other industry applications. But language is hard for computers to understand. The language we use to communicate is not just a collection of symbols—it is rich in context beyond mere words and makes it very difficult for machines to interpret. As luck would have it, much of the information that organizations can get insights from or use to automate functions (such as product reviews in social media posts, customer call center recordings, physicians’ notes, contracts ...