Currently on the CommunicAsia exhibition floor: a Canadian startup that aims to democratize data query for business data via a natural language interface that allows non-techy execs to speak queries rather than type them.
Kelly Cherniwchan, CEO and co-founder of Chata, explained that many executives feel left out and unable to do their tasks because of technological barriers. The status quo is that either you get someone in to write an SQL query to run, or you export all the data into Excel and go after that.
What Chata is trying to do is alleviate that frustration by putting in a natural language interface for the executive to type in queries naturally. This could be as simple as, “Show me all the product sales from the state of Texas”, or something more complicated like, “Show me the average scan time per physician”.
Cherniwchan quoted McKinsey research saying that one-fith of an employee’s time is spent searching for data and for knowledge workers. An IDC report put that figure at between 30% to 40%.
Behind the scenes, Chata uses a natural language model and computational linguistics to determine what it is the user wants.
“If you think about the market share of the whole natural language query, data query and analysis is just a small part. But we benchmark ourselves against [IBM] Watson and others,” Cherniwchan said.
The system will understand complex requests such as, “show me the average read time per physician per indication”. You can flip it and ask, “show me the average read time per indication per physician,” and get the desired answer.
Then there are query chains. A user can ask for the average scan time, then chain the query and ask for only data from patients over 50. In a future version, the AI can differentiate between chain queries and new queries.
A more complicated example is one oil and gas client who wanted to be able to ask, “show me all projects that are behind schedule”. That does not mean anything to the natural language AI – it needs to be broken down through what Cherniwchan calls natural language macros. What the above request is really asking is, “show me all the numbers where the requested completion dates are before the scheduled start dates”.
Cherniwchan explained how his AI was learning from the queries. By studying the queries, the system will learn how human brains work and how they structure their queries in each industry and how users like to chain queries.
The company started with the idea of building a chatbot for people, and capitalizing on the conversational UI to cater for the next generation of executives who did not grow up with a screen and keyboard, but with a mobile phone. But along the way it became clear that there was a huge market in building a user interface for small to medium sized businesses to access their data in a way that is better than the Microsoft Excel model. The first iteration was bootstrapped on API.AI (now bought by Google) before they went on to write their own system.
Some of the exciting projects Chata has in the pipeline include one with Quickbooks to allow natural language interface for a user to access their own financial data, and another with a bank to provide natural language queries on their bank accounts and debit card spending.