Teaching search engines to know what you need

Information scientists are figuring out how to make search engines better at listening, so they can give us search results before we even ask.

Quizzing Siri or Alexa on the weather forecast or latest football results is common these days. But if we’re going to have fluent and intuitive conversations with future search engines, they will need to be built differently.

This challenge has captivated Professor Mark Sanderson, Director of the Enabling Capability Platform for Information and Systems (Engineering), and his team at RMIT University.

They’re trying to better understand how we talk about information to each other, to develop search engines that can do the same. The next step is building a new kind of search engine—one that listens in on people.

PhD student Johanne Trippas from RMIT and Dr Hideo Joho from the University of Tsukuba, Japan, have been analysing dialogues in which people are planning trips together, talking about bookings and potential holiday activities.

By breaking conversations down into ‘information needs’, researchers can establish the technological features needed for a next-generation search engine.

“A really clever search engine might listen to our conversation and, for example, throw up a video that I’ve been talking about,” Mark says.

Instead of choosing menu options and typing in queries, this work on conversational search engines is taking us to a future where we’ll be able to have a casual chat with a computer.

“Obviously, there are privacy issues,” Mark says.

“The idea of a computer just continually listening sounds a bit scary, but the initial question is—can we make this work or not?” 

For more information:

RMIT University Research & Innovation
James Giggacher
+61 03 9925 4143