(Un)likelihood Training for Interpretable Embedding
This work addresses the problem of interpretable and predictable video search for applications like ad-hoc retrieval, though it appears incremental as it builds on existing cross-modal representation learning methods.
The paper tackles the challenges of black-box training and dataset bias in cross-modal representation learning for video understanding by proposing likelihood and unlikelihood training objectives to interpret embeddings and address label sparsity, resulting in a new encoder-decoder network that outperforms state-of-the-art retrieval models on TRECVid and MSR-VTT datasets with statistically significant margins.
Cross-modal representation learning has become a new normal for bridging the semantic gap between text and visual data. Learning modality agnostic representations in a continuous latent space, however, is often treated as a black-box data-driven training process. It is well-known that the effectiveness of representation learning depends heavily on the quality and scale of training data. For video representation learning, having a complete set of labels that annotate the full spectrum of video content for training is highly difficult if not impossible. These issues, black-box training and dataset bias, make representation learning practically challenging to be deployed for video understanding due to unexplainable and unpredictable results. In this paper, we propose two novel training objectives, likelihood and unlikelihood functions, to unroll semantics behind embeddings while addressing the label sparsity problem in training. The likelihood training aims to interpret semantics of embeddings beyond training labels, while the unlikelihood training leverages prior knowledge for regularization to ensure semantically coherent interpretation. With both training objectives, a new encoder-decoder network, which learns interpretable cross-modal representation, is proposed for ad-hoc video search. Extensive experiments on TRECVid and MSR-VTT datasets show the proposed network outperforms several state-of-the-art retrieval models with a statistically significant performance margin.