This document presents a method for ranking the relevance of comments on YouTube videos. It involves a two-stage process: first classifying comments as relevant or noisy, then scoring the relevant comments based on topics extracted from the comments and related online resources about the video. The authors describe their classification of comments using linguistic features and a neural network. They also explain three approaches to relevance scoring using topics extracted from comments, Wikipedia articles, and lyrics. The results indicate the method produces a different comment ranking than YouTube, but more evaluation is needed to properly assess the impact of the relevance measure.