Topic Modeling in SE

Lots of software engineering involves processing unstructured text, e.g., from issue discussions, documentation sites, app reviews. Topic modeling is a technique for capturing the latent (hidden) topics of a document that has been widely applied in SE.

Learning Outcomes

• overview knowledge of the way statistical topic models work.
• analyze a topic model and understand its benefits and drawbacks.

Lecture Notes

• LDA tutorial (Swapnil Hingmire)

Required Readings

• Campbell, Hindle, Stroulia, Latent Dirichlet Allocation: Extracting Topics from Software Engineering Data
• Agrawal, What is Wrong with Topic Modeling

Activity

• Intro to BertTopic - this is a Colab notebook. Please go through it ahead of class.

Optional Readings and Activities

• compare this to the Novielli work on sentiment analysis.
• Blei, Ng, Jordan, LDA. Classic paper introducing Latent Dirichlet Allocation.
• Barua, What Are Developers Talking About? An Analysis of Topics and Trends in Stack Overflow