CXP 17: Leveraging Natural Language Understanding for Self-Service Chatbots


Last week’s CXP series post covered using CX Designer for self-service application creation and today we are going to continue the series by covering the additional capabilities that take our chatbot creation platform to a new level.cxp-app-lifecycle-management

Interactive Text Response (ITR) is so much more than simply recasting traditional IVR interactions into a text channel.   In IVR, the initial action from the user is to place a phone call – an act that conveys no content – so the system initiates the dialogue.  The applications’ designs tend to focus on tightly constrained responses to specific prompts.  In a chatbot environment, however, the user is playing with the white pieces in the game of dialogue chess; she moves first.  That first move could contain rich, open-ended content, and the challenge to understand it is a weighty one.

To meet that challenge, CXP 17 brings to the chessboard a toolkit that takes the understanding of language far beyond keywords.  A complete Natural Language Understanding (NLU) engine is now beneath the hood, empowering the application to see user input not as characters, but as communication – with the attached semantic content (meaning), and a full grasp of the syntactic relationships (grammar).  This allows the application developer to leverage a powerful reasoning component that explicitly creates a mapping of user input content to user intent.  With its grammatical power, CXP 17 can make the fine distinction between “How do I charge my battery” and “How much do you charge for an extra battery?” while leaving the ways in which user intents are determined completely transparent to the developer.  A description of an intent can be locked down to the specific pairing of an action and its object (like charge and battery) while recognizing that semantic roles remain the same even when the sentence’s order moves around, as in the sentence “How does my battery get charged?”  This is all accomplished while enabling you to get your bot running while you do not yet have large amounts of training data.

Chatbots and natural language understanding

Because the reasoning engine in CXP uses semantic abstractions, the logical expressions are interlingual – write them for a bot that interacts in English, and they will also recognize user intents expressed in Spanish.  These semantic abstractions also automatically account for inflections, synonyms, and categories of items like food and clothing.  Furthermore, extracting common data types such as date, time, or money is now a breeze. Whether your customers are expressing a date as “next Wednesday”, “in 7 days”, “June 28th”, or “6/28”, Aspect NLU returns 20170628 for you. And because this channel frequently involves modes of expression that don’t strictly adhere to the dictionary, unknown words, i.e. typos are automatically replaced with the appropriate candidates so that our ability to understand is not thwarted by the habits of trendy millennials.

For a brief example of how to get started building a bot using NLU, see Tobias Goebel’s demonstration video showing how easy it is with the CX Designer interface.


Lisa Michaud

Lisa Michaud is the Director of Natural Language Processing (NLP) at Aspect. She has been centrally involved in the integration of NLP components into Aspect’s product suite for customer engagement and the architecting of our Interactive Text Response (chatbot) technology. She has 20 years of research experience in the field of Natural Language Processing / Computational Linguistics and pursues diverse interests in user modeling, dialogue, parsing, generation, and the analysis of non-grammatical text.She holds a PhD in Computer Science and has been published in multiple international journals, workshops, and conferences in the fields of user-adaptive interaction and Computational Linguistics.