I don’t envy newcomers into the Voice AI arena. It’s so hard to get around with so many acronyms and terms; machine learning, deep learning, neural network, AI, NLU, NLP, NLG and more. Sometimes it seems that the new rapidly emerging technologies grow faster than the actual Voice AI market.
So, let’s “Begin at the beginning" (― Lewis Carroll, Alice in Wonderland); two important terms in the world of Voice AI that get easily confused are Natural Language Processing (NLP) and Natural Language Understanding (NLU).
Natural Language Processing (NLP) is the general term used to describe systems that help machines understand and analyze all the complexities of human language.
Natural Language Understanding (NLU) is a subset of NLP, a more advanced application of language processing. The NLU attempts to handle unstructured inputs by semantically analyzing the user’s intention and tone; the process through which computers understand language.
Since personalization and speed of service is key, one of the more complex problems in the voice AI industry today is making sure machines will be able to understand users even when they improvise and talk in the most natural (day to day), unpredicted way.
Although it will be possible in the future, today even Siri and Alexa are not yet at the point where the user can talk freely with 100% accuracy of speech, tone and intention recognition.
So, until the machines catch up with human speech complexity, we believe that the next best thing is a Guided-NLU (Guided - Natural Language Understanding) engine that applies a domain-centric approach to speech processing and enables users to use natural language within a pre-determined, pre-defined set of boundaries and terms, as guided by the domain’s ontology.
Guided-NLU engines should be built with one main principle in mind- each industry has its own common jargon. When calling our credit card company, we will probably use words related to our account and charges (e.g. Please check my balance, My CVV number is worn out- what can I do?) and not words related to what toppings we might want on our pizzas 😊
Therefore, by guiding a voice recognition engine to focus on a pre-determined set of words, terms and slang, we can increase recognition accuracy dramatically. By using Guided-NLU we can increase the automatic handling rates, while still making sure callers get to enjoy free speech and conversation with a machine that is completely tuned in to their specific needs.
Learn more about our Voice.AI solutions for enterprises.