43 - Reinforced Video Captioning with Entailment Rewards, with Ramakanth and Mohit - podcast episode cover

43 - Reinforced Video Captioning with Entailment Rewards, with Ramakanth and Mohit

Dec 04, 201748 min
--:--
--:--
Download Metacast podcast app
Listen to this episode in Metacast mobile app
Don't just listen to podcasts. Learn from them with transcripts, summaries, and chapters for every episode. Skim, search, and bookmark insights. Learn more

Episode description

EMNLP 2017 paper by Ramakanth Pasunuru and Mohit Bansal Ram and Mohit join us to talk about their work, which uses reinforcement learning to improve performance on a video captioning task. They directly optimize CIDEr, a popular image/video captioning metric, using policy gradient methods, then use a modified version of CIDEr that penalizes the model when it fails to produce a caption that is _entailed_ by the correct caption. In our discussion, we hit on what video captioning is, what typical models look like for this task, and how the entailment-based reward function is similar to other attempts to be smart about handling paraphrases when evaluating or training language generation models. Unfortunately, due to some technical issues, the audio recording is a little worse than usual for this episode. Our apologies. https://www.semanticscholar.org/paper/Reinforced-Video-Captioning-with-Entailment-Reward-Pasunuru-Bansal/0d11977afa1a6ce90dc3b1f26694492c2ab04773
For the best experience, listen in Metacast app for iOS or Android