About This Event
Valuable data that could be used for business insights and decision-making is often only available in free form text fields. This presentation will review one method of converting text to more usable information by using Latent Dirichlet Allocation (LDA) topic modeling. This discussion will include an overview of where this method was applied to voicemail transcripts to quantify the most common reasons claims participants call claim representatives, with the results helping to prioritize IT investment in building out self-service functionality.
What is Latent Dirichlet Allocation?
- Methodology
- Model Outputs
- Performance Metrics
Applying LDA in Practice
- Project Overview
- Data Pre-processing
- LDA Model Set-up & Evaluation
- Topic Results
This webinar is free. We welcome all iCAS members, CAS members and non-members to attend.
Denise Christophel, MS, CSPA, CPCU
Denise Christophel is a Director of Advanced Analytics at Sentry Insurance. She leads a team of talented individuals who provide analytics and business intelligence solutions for commercial lines and several other areas within Sentry. Her educational background includes a MS in Statistics from Oklahoma State University and a BA in Psychology from University of Northern Iowa. She has also earned the Certified Specialist in Predictive Analytics (CSPA) credential through the CAS Institute.
