CPSC 503 - Pedagogical Project
This repository holds all learning material prepared for the pedagogical project as part of the CSPC 503 (Spring 2019) course at UBC.
Latent Dirichlet Allocation (LDA)
Latent Dirichlet Allocation model (LDA; Blei et al. 2003) is a hierarchical Bayesian approach to topic modelling that describes a generative process of document creation. The goal of LDA is to infer topics as latent variables from the observed distribution of words in each document.
Why learn LDA?
- Within the list of topic modelling methods, LDA is the most widely used one, especially in research disciplines relying heavily on textual data to extract information.
- In the computer science literature, the research paper by Blei et al. (2003) that proposed LDA for the first time is one of the most cited papers in machine learning.
Main Content
In this repo, you can find:
- the lecture slides,
- the final report explaining in detail the project goals and content,
- a simple demonstration of LDA in R language, and
- an assignment designed to assess the learning experience of students
Resources
There are many great resources online to learn and apply LDA. Needless to say, the best resource to start with is the Blei et al. (2003) article that introduced the LDA algorithm to the literature.
Below is a list of other resources that have been greatly helpful in preparing the material within this repository.
For understanding LDA:
-
Video lectures by David Blei at Cambridge University in 2009
-
Topic modelling sections of the CPSC 540: Machine Learning slides by Mark Schmidt
-
Introduction to Latent Dirichlet Allocation on Edwin Chen's blog for a fun and intuitive explanation of LDA
For applying LDA:
- Text Mining with R by Julia Silge and David Robinson