Papers with ablative studies
Video Question Answering with Phrases via Semantic Roles (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing VidQA evaluation metrics limit the models’ application scenario to a single-word answer or selecting a phrase from a fixed set of phrases. |
| Approach: | They propose to leverage video descriptions to mask out certain phrases to enable evaluation of answer phrases. |
| Outcome: | The proposed model reduces the influence of language bias on VidQA datasets by retrieving a video having a different answer for the same question. |