Next-Generation Conversational AI for Database Access:
The Technology Behind AutoQL
How innovation in conversational AI is revolutionizing the way today's leading businesses access and leverage their data
Chapter 4
Going Beyond Intent Classification
Let’s start with an example of how humans typically communicate by considering the sentence “Who owes me?” This is the type of question that humans are equipped to answer, but difficult for computers to understand.
“Who owes me?” is a question asked in context and, though it might refer to a specific piece of information, there’s a lot of ambiguity around what that specific information might be.
In other words, the person asking the question has an idea of the information they’re looking for, but they haven’t clearly stated what is owed. Maybe they need you to tell them who has yet to pay them back for dinner last night, or maybe they’re asking which of their clients is late in paying their monthly invoices. The entities that would typically be used to classify the intent of this statement are not present.
Intent classifiers don’t offer the flexibility, accuracy, or opportunity to access and explore data conversationally.
Humans can deduce intent by factoring in real-life context (what the intention might be while talking to a friend versus speaking with the accountant).
An intent classifier might have some level of context if it’s built for a specific purpose, say, consumer banking, but because the intent classifier can only match a phrase to an intent that it already knows exists, this limits its flexibility. Unless every single possible intent is built into the intent classification system, there will always be gaps in the machine’s understanding and ability to respond to what a human is really asking.
Another drawback of intent classifiers is limited NLU power. The AI only has to understand human language insofar as it can apply an intent category to statements that match its pre-defined list of intents. That means users still need to make sure that they ask questions in words that the AI is likely to understand.
This can sometimes feel like a tedious, even frustrating, game as users attempt to guess the words that the computer knows, while the system continuously returns the old refrain: “I’m sorry, I don’t understand what you’re asking for.”
At Chata, we see a gap in AI technology aimed at making data access faster and more intuitive, particularly when it comes to enterprise-grade databases. While intent classifiers have been built to provide better experiences in customer service and marketing channels, they don’t offer the flexibility or opportunity that is necessary for accessing data conversationally.
AutoQL goes beyond intent classification: our system understands full natural language statements and dynamically generates database query language.
This is because conversational AI built specifically for database access requires a level of complexity that can’t be achieved through the limited application of intent classifiers: it’s simply too labor and time intensive to create the massive number of intent categories database users might be interested in exploring.
Behind this limitation is the sheer volume of training data that would be needed to encapsulate the scope of an entire database, the business logic associated with that database, as well as all the types of questions users might ask about their data.
To fill this gap, it’s time to go beyond intent classification and build AI specifically for conversational data experiences.
