This document discusses mining emotions from user comments on short films. It presents an approach that creates an emotion vector for each short film based on extracting terms from user comments on YouTube and associating them with emotions from the NRC Emotion Lexicon. It then compares the cosine similarity between emotion vectors built from expert judgments and those built using Amazon Mechanical Turk workers or automatically from YouTube comments. The goal is to determine if crowdsourcing or YouTube comments can accurately extract emotions expressed in reviews of short films.