Editor’s note: Tom is a speaker for ODSC East 2019 this April 30-May 3! Be sure to attend his talk, “Crisis Intervention and Saving Lives with Natural Language Processing & Predictive Analytics.”
The field of predictive modeling is still young, but one of the most effective trailblazers is Crisis Text Line, a free, anonymous 24/7 text-based crisis intervention system that aims to mitigate crises by connecting people to counselors who are trained to cool down hot moments. The service specializes in short-term crisis intervention and has handled more than 2 million conversations to date with more than 65 million texts exchanged.
With a team of over 4,000 crisis counselors, Crisis Text Line takes a human-first approach that uses data to supplement a counselor’s conversation and improve the outcome of a crisis. The organization was founded with a focus on using data to help people through tough times and has built an impressive set of mental health data to inform interactions.
Crisis Text Line uses Periscope Data to analyze text conversations in real time. They use natural language processing and machine learning to pull insights from their rich data set and identify keywords in texts to help steer a counselor toward a safe resolution. Later, the second phase of this process utilizes a large community of professional counselors to analyze conversations based on common keywords and tags to help assess trends and train counselors to have high-quality conversations with texters.
This innovative approach to predictive modeling allows Crisis Text Line to detect keywords that identify and predict trends in real time. This augments the counselors’ abilities and gives them the peace of mind that they aren’t alone. While the advanced analytics are valued for their assistance, it is important to the organization that they don’t replace human volunteers with automated responses.
[Related article: Machine Learning and Compression Systems in Communications and Healthcare]
The Crisis Text Line data team uses Periscope Data to conduct this complex analysis and quickly visualize the results. In the near future, the team plans to set up a self-service data environment that will empower counselors to access information without help from the data team. This setup would give counselors quicker access to data and ultimately lead to better-informed conversations with texters. Often, the end users have difficulty predicting the needs of texters ahead of time, so a data tool that relies on upfront modeling is ineffective. An agile data environment like Periscope’s allows the team of counselors to find answers on their own.
[Related article: Introduction to Clinical Natural Language Processing]
Periscope Data has helped the Crisis Text Line team scale significantly in the past year and there are plans to more than double the number of volunteers and conversations again over the next two years. By empowering more end users to access advanced analytics dashboards, the service can rely on counselors to find more answers on their own and give texters a better experience as the team grows.
