Video to Skill Tagging using Transcripts under Weak Supervision

Zhe Cui, Shivani Rao

31st Conference on Neural Information Processing Systems (NIPS 2017) Workshop: Learning with Limited Labeled Data: Weak Supervision and Beyond |

Delivering micro-content (small pieces of learning materials) from a course catalog has become important as it helps learners engage quickly with a learning platform on a need-to-know period. In order to recommend videos that can teach skills relevant to a learner, we propose an approach for tagging videos with skills using transcripts. Since there is no labeled data to train a video-to-skill model directly, we leverage a bunch of techniques that fall under the umbrella of weak supervision. We evaluate our approach using offline metrics like Precision, Recall, F-1 score and online metrics via large-scale A/B testing.