Detecting Concrete Visual Tokens for Multimodal Machine Translation
This work addresses visual grounding challenges in multimodal machine translation, but it appears incremental as it builds on existing architectures and datasets.
The paper tackled the problem of detecting and selecting visually-grounded tokens for masking in multimodal machine translation, introducing new detection and selection methods that improved performance and visual context usage over a baseline model.
The challenge of visual grounding and masking in multimodal machine translation (MMT) systems has encouraged varying approaches to the detection and selection of visually-grounded text tokens for masking. We introduce new methods for detection of visually and contextually relevant (concrete) tokens from source sentences, including detection with natural language processing (NLP), detection with object detection, and a joint detection-verification technique. We also introduce new methods for selection of detected tokens, including shortest $n$ tokens, longest $n$ tokens, and all detected concrete tokens. We utilize the GRAM MMT architecture to train models against synthetically collated multimodal datasets of source images with masked sentences, showing performance improvements and improved usage of visual context during translation tasks over the baseline model.