End-to-end Concept Word Detection for Video Captioning, Retrieval, and Question Answering
This work addresses the need for improved semantic priors in video captioning, retrieval, and question answering, offering an incremental enhancement by integrating a novel detector with existing models.
The authors tackled the problem of generating semantic priors for video-to-language tasks by proposing a trainable concept word detector that does not require external knowledge, achieving best accuracies in three out of four LSMDC 2016 tasks, such as fill-in-the-blank and movie retrieval.
We propose a high-level concept word detector that can be integrated with any video-to-language models. It takes a video as input and generates a list of concept words as useful semantic priors for language generation models. The proposed word detector has two important properties. First, it does not require any external knowledge sources for training. Second, the proposed word detector is trainable in an end-to-end manner jointly with any video-to-language models. To maximize the values of detected words, we also develop a semantic attention mechanism that selectively focuses on the detected concept words and fuse them with the word encoding and decoding in the language model. In order to demonstrate that the proposed approach indeed improves the performance of multiple video-to-language tasks, we participate in four tasks of LSMDC 2016. Our approach achieves the best accuracies in three of them, including fill-in-the-blank, multiple-choice test, and movie retrieval. We also attain comparable performance for the other task, movie description.