Luminoso’s new ‘Compass’ tool navigates through online chatter

A snapshot of the concept cloud taken from Luminoso's "Dashboard" product - the cloud showcases Luminoso's ability to understand and analyze emojis as language - giving meaning to the symbols that are plastered all over social media and the internet, in this case, as they relate to "Boston Snow."
A snapshot of the concept cloud taken from Luminoso's "Dashboard" product - the cloud showcases Luminoso's ability to understand and analyze emojis as language - giving meaning to the symbols that are plastered all over social media and the internet, in this case, as they relate to "Boston Snow."

The sound of social media is more cacophony than symphony. But if you have the right tools to listen, you can cut through the noise. Such is the theory behind Compass, the new platform launched Wednesday by Luminoso, the social analytics firm in Cambridge with roots in the MIT Media Lab.

Compass’ technology builds on the ConceptNet tools first developed by MIT in the ’90s to teach computers natural language patterns used by humans. The new tool sorts through the flood of muddled information from Twitter, Facebook, and Google+ and uses machine learning to process text (and even emojis) in real-time to sift out trends. It can make sense of words that are contextually relevant to a subject, telling the difference, for example, between Apple computers and actual apples, or the Oracle software firm and your run-of-the-mill seers.

“When people talk about things that they’re passionate about — be it products or food, all the way down to emotions about events — they use really creative language, and that language isn’t usually predictable,” said Catherine Havasi, Luminoso co-founder and chief executive.

Havasi said Compass will identify connections that would go unseen in traditional keyword searches on social media. A query for “Boston” and “snow” using Compass would also identify stories about MBTA delays or wandering Yetis, she said.

She said the tool can be used by companies during product launches to detect software glitches or negative feedback and respond to issues before they go viral. Compass also has a component used by news organizations to track trends on Twitter before they gain critical mass.

“When you approach text data it can be different from when you approach structured data,” said Fern Halper, the research director for advanced analytics at The Data Warehousing Institute. “Their approach is very interesting. They’re applying machine learning to learn about different concepts and themes and developing their own set of metrics.”

Luminoso tested a prototype of Compass during last summer’s FIFA World Cup, when it partnered with Sony to slice and dice the social media chaos happening around the planet in six different languages. “They couldn’t have a core of moderators updating the social media queries for every nickname of every player,” Havasi said. “Much of what we’re replacing is people reading Twitter.”

For its launch, Compass has expanded its reach and can now interpret text in eight languages, and emoji. “For us emoji are essentially words—they’re different ways of expressing different kinds of sentiments,” Havasi said. “You really need to be able to understand that there’s a lot of content and meaning that is associated with them culturally, and they carry different meanings in different circumstances.”

Janelle Nanos can be reached at janelle.nanos@globe.com. Follow her on Twitter @janellenanos.
Follow Janelle on Twitter