Learn How Tech Startup Crisis Text Line is Utilizing Big Data and Machine Leaning to Save Lives – All Via Text Messages

This year’s Strata Data Conference in San Jose, California, had keynote speakers Nancy Lublin and Bob Filbin centerstage to discuss how their tech startup Crisis Text Line utilizes big data and machine learning to save lives via text messaging.

IMG_1369_editThis year’s Strata Data Conference in San Jose, California, had keynote speakers Nancy Lublin and Bob Filbin centerstage to discuss how their tech startup Crisis Text Line utilizes big data and machine learning to save lives via text messaging. Crisis Text Line is a free, 24/7, confidential text message service for people in crisis – particularly those at risk of suicide.

Why text messages? They provide a more convenient way to communicate that is private – no one can overhear or know what you are sending. They’re easy and fast access to help.

In the past five years, Crisis Text Line has answered more than 65 million text messages. Seventy-five percent of its users are under the age of 25 and ten percent are under the age of 13. Area codes with the ten percent lowest social economic status utilize around nineteen percent of the total volume.

With all that data, the big challenge for the organization was to find a way to prioritize and respond to the texters at imminent risk first. “Crisis help should be answered based on severity, not chronology,” Filbin said.

The initial phase included list of fifty key words commonly associated with suicidal risk. This resulted in the organization being able to identity a little over fifty percent of texters who were at imminent risk with a response rate of thirty-nine seconds.

In the second phase, Crisis Text Line was able to take the large amount of data from the sixty-five million texters and build a deep neural net utilizing machine learning. This increased the initial keyword list to ten-thousand unigrams, bigrams and trigrams that indicate suicide. What was discovered is that some words posed greater risk than the word suicide. Examples include the word military, which is twice as likely to indicate imminent risk. Common household drugs names such as Aspirin and ibuprofen and drugstore names were also at the top of list. Symbols were also part of this list; the crime face emoticon indicated a person is eleven times more likely to be in imminent risk.

The implementation of machine learning allowed Crisis Text Line to identify eighty-five percent of the most severe cases by the first text message, allowing the best opportunity to help those individuals in the heat of the moment and affect positive outcomes. This is an excellent example of how emerging technology is being used to benefit society and save lives.

Click here to learn more about new and evolving tech in the third issue of CompTIAWorld.

Read More from the CompTIA Blog

Leave a Comment