Unimodal and Multimodal Representation Training for Relation Extraction
This work provides insights into modality contributions for relation extraction, which is important for document understanding applications, though it is incremental in nature.
The study investigated the relative importance of text, layout, and visual modalities for relation extraction in visually rich documents, finding that a bimodal text and layout approach achieved the best performance (F1=0.684), with text being the most predictive single modality.
Multimodal integration of text, layout and visual information has achieved SOTA results in visually rich document understanding (VrDU) tasks, including relation extraction (RE). However, despite its importance, evaluation of the relative predictive capacity of these modalities is less prevalent. Here, we demonstrate the value of shared representations for RE tasks by conducting experiments in which each data type is iteratively excluded during training. In addition, text and layout data are evaluated in isolation. While a bimodal text and layout approach performs best (F1=0.684), we show that text is the most important single predictor of entity relations. Additionally, layout geometry is highly predictive and may even be a feasible unimodal approach. Despite being less effective, we highlight circumstances where visual information can bolster performance. In total, our results demonstrate the efficacy of training joint representations for RE.