We often come across the term understanding in our work – which is to make machines interpret commonly spoken natural language. Why is understanding important in the context of natural language-handling by machines aka Natural Language Understanding (NLU)?

To us at AUI Systems, Understanding has a specific connotation related to meaning and context. It also represents the potential unlocking of massive business value that currently appears beyond the reach of statistical or Machine Learning-powered NLP systems.

Understanding is at the heart of NLU. When humans understand a sentence – we understand the meanings of the individual words and the context in which they are written or spoken. We draw context from personal knowledge and experience stored in our brains. If we are confused, we look up a subject matter reference or consult with someone possessing the relevant experience. In short, we always reference a knowledge source to obtain context and understanding. We naturally place the understood information in our internal knowledge base, our brain, in the right contextual and temporal setting – to be drawn upon at some future time.

So why is understanding in machines so important to business value? When a machine really understands the entire content in an earnings call transcript or a customer contact center conversation, it can extract the what, when, where, why, and who within the transcript or customer conversation.

It is thus able to convert unstructured information into structured knowledge that can be accessed interactively by humans using natural language conversations. This represents a massive jump in productivity for business analysts who earlier spent considerable time reading documents to unearth insights. Cross referencing of knowledge across time periods is another huge plus.

Here is a practical example: we quote an excerpt from an Apple, Inc. earnings call transcript in 2019: “We accomplished these results despite strong headwinds from foreign exchange which impacted our top-line growth rate by 300 basis points compared to a year ago”. To deliver value we believe NLU must be able to understand here that overall revenue growth for the relevant quarter was impacted negatively by 300 basis points on a year-on-year basis due to a foreign exchange impact. This understanding is possible only when the machine unambiguously grasps the meaning of all the words in the sentence, the syntax, and the context of this sentence within the overall transcript. Specifically, words such as headwinds, foreign exchange, basis points and others and their relevance in finance must be part of the machine’s knowledge system – hence are understood and converted to semantic knowledge.

When this level of understanding occurs, tremendous business value is unleashed. For example, productivity jumps when analysts can obtain qualitative and quantitative insights from earnings calls transcripts at a keystroke. Or when NLU powers contact center conversations to resolution without human intervention. AUI Systems has developed an NLU engine that delivers real understanding and is available for use.