Unsupervised Discovery of Multimodal Links in Multi-image, Multi-sentence Documents
This addresses the challenge of linking images and text in web documents for applications like content analysis, but it is incremental as it builds on existing multimodal methods.
The paper tackled the problem of discovering image-sentence relationships in multimodal documents without explicit annotations, and found that a structured training objective based on co-occurrence in documents can predict specific links, as tested on seven datasets of varying difficulty.
Images and text co-occur constantly on the web, but explicit links between images and sentences (or other intra-document textual units) are often not present. We present algorithms that discover image-sentence relationships without relying on explicit multimodal annotation in training. We experiment on seven datasets of varying difficulty, ranging from documents consisting of groups of images captioned post hoc by crowdworkers to naturally-occurring user-generated multimodal documents. We find that a structured training objective based on identifying whether collections of images and sentences co-occur in documents can suffice to predict links between specific sentences and specific images within the same document at test time.