Watch talk ➟ http://bit.ly/1SGlQqj
Interested in what your customers (or their customers) are talking about, and how that’s changing over time? The most popular 'topic trending' analyses depend on the investigation of how keyword (or term) usage change over time. Using keyword trends to index topic trends is highly suitable for short-form text, such as search terms, hashtags, or tweets. However, exploring trends in topic prevalence across longer, more free-form texts, such as call center telephone transcripts, is better served by grouping together topically-related words that co-occur. In this talk, we'll show an example of using Latent Dirichlet Allocation (LDA; Blei, Ng, & Jordan, 2003) in R to group words in call center transcripts into multiword topics over which we explore trends. We’ll also demonstrate the how the use of Structural Topic Modeling (Roberts, Stewart, Tingley, & Airoldi, 2013) can aid in further investigation of how document-level covariates (in this case, additional call- or caller-based characteristics) can affect topic prevalence and topic trends.
Data Science Popup Austin: Using lda and Structural Topic Modeling to Explore Trending Topics in a Call Center
1. DATA
SCIENCE
POP UP
AUSTIN
Using LDA and Structural Topic Modeling to
Explore Trending Topics in a Call Center
Jordana Heller
Data Scientist, Mattersight
jheller
4. Lightning Talk:
Using LDA and Structural Topic Modeling to
Explore Trending Topics in a Call Center
Jordana Heller @jheller
Data Science Pop-up Austin, April 13, 2016