The Musigma webinar series

2020-02: Topic Modeling Using LDA

February 4, 2020

Speaker: Bret Shroyer, cognalysis

It’s a common challenge for actuaries and data scientists – “How can we transform free-form text into quantitative data for modeling or reporting?”

One popular and intuitive method is Topic Modeling: Assigning topic categories free-form text using an algorithm, such as Latent Dirichlet Allocation. This webinar is aimed at actuaries with limited prior exposure to text modeling, with the goal of teaching how to perform topic modeling, what sorts of data is best suited for topic modeling, and how to understand the output of topic models.

Presented by Bret Shroyer, FCAS of Cognalysis. Bret had previously presented this topic to actuaries at the CAS annual meeting in November 2019, and at the MinnCAS assembly of actuaries in December 2019. Bret will present not only the theory and process of topic modeling, but will also walk the audience through a live example of topic modeling using both Excel and R. There will be opportunities for audience participation, including live polls. The Excel workbooks, R scripts, and data will be provided to the audience for their use.

Presented by Cognalysis, the MuSigma Webinar Series is an excellent opportunity for actuaries and insurance professionals to engage in free, interactive, and relevant content designed to provide quality continuing education.

You can download the Excel files, the R code, and the Wine Review data at