At the most basic level, computers communicate using binarycode—zeros and ones. To get from bits and bytes to conversational English—orany other human language—requires complex technological development. It’s beena long time coming, but the area of artificial intelligence (AI) called naturallanguage processing, or NLP, has finally advanced to the point where apps canstand in for human agents in a multitude of ways, many of which are cropping upin eLearning and performance support tools. This article describes a few waysL&D teams are already using NLP in eLearning, but it is far from a completelist.
What is NLP?
NLP technologies parse complex human language, using algorithmsto identify parts of speech, analyze syntax, and figure out the function ofeach word in a phrase or sentence. They can then decipher the meaning of acomment or request and respond appropriately. NLP draws on other areas of AI,such as machinelearning and deeplearning, technologies that enable computers to “learn” to performtasks that humans typically perform, including recognizing speech, learningspeech patterns and variations, and formulating responses that mimic humanconversation. These technologies make it possible for NLP-powered apps to workfluidly, rather than being limited to a predetermined set of responses. The abilityof AI-powered technology to learn and build on experience and inputs has moved computersfrom the era of programmed inputs into what the Future Today Institutecalls “cognitive”computing—an age where computers problem-solve and apply reasoning.
A person using an NLP-based technology might speak to theapp or device, type, or input text in another way: When a learner asks Siri aquestion, instructs her smart speaker to play music, or dictates a text oremail, she’s using NLP. Alternatively, a learner can type into an email ortexting app, and an NLP-powered feature might suggest word completions. Emailprograms that suggest a response to an incoming message are also using NLP.
In many languages, words can have different meaningsdepending on the context in which they are used. This “structural ambiguity”presented enormous challenges in the development of NLP algorithms. MadlyAmbiguous, an online game where players attempt to trip up the computer,shows—and explains—how one NLP algorithm deals with structural ambiguity by“learning” all possible meanings of a word, then analyzing the other words in aphrase to tease out the correct meaning for a specific context.
NLP in eLearning
With the technology integrated into many apps and tools, learnersmight not realize how frequently they are using natural language processing ineLearning.
For instance, any chatbot-based performance support toolrelies on NLP—and chatbots are increasingly adept conversationalists. Peopledon’t always know (or care) that they are interacting with a chatbot: Manystudents using chatbot-based virtual assistant at Georgia Tech didn’t evenrealize that “she” wasn’t human, according to ArtificialIntelligence Across Industries: Where Does L&D Fit?, a Guild Research report by JaneBozarth.
In the eLearning context, chatbots might:
- Provide first-line customer service or technicalsupport, responding to common, routine queries and escalating more complexproblems to a human agent
- Guide new employees through an onboardingprocess, alerting them to tasks they need to complete, reminding them ofdeadlines, and sending them information or links to forms
- Coachlearners between face-to-face or virtual classroom sessions, reviewingmaterial covered or asking quiz questions
- Check in with trainees, asking them to reflecton skills learned or observations
- Provide spaced practice and drills to supplementor reinforceeLearning or face-to-face training
- Offer performancesupport by answering questions, linking employees to information ordocuments, assisting with basic tasks, or stepping them through seldom-usedprocesses
Voice-based virtual assistants are not restricted tochatbots on mobile devices. Smart devices, reportedlyused by more than 50 million American adults, are already performingroutine office tasks. In some companies, voice-activated assistants scheduleconference calls and connect participants, for example.
NLP-powered apps can create text as well as interpret it.Within L&D, NLP can be used to:
- Generate short content based on keywords, streamliningthe content creation process
- Extract keywords from existing content, categorizingcontent, facilitating searches, and helping identify relevant eLearning anddeliver it to learners
- Summarize long texts by identifying key themesor ideas, aiding in chunking of long content and targeting of content tolearners
- Determine whether a passage, such as a commenton an eLearning “smile sheet,” indicates a positive or negative sentiment, a usefulway to gauge learners’ responses to eLearning
- Identify patterns or themes across multipletexts to categorize, curate,and target content
- Translate between languages, enabling cross-culturalconversation and document sharing among employees at global companies
- Animate virtual trainers and make branchingscenarios in eLearning more dynamic and engaging
- Create voice-activated performance support toolsthat allow employees to get assistancein the workflow
