Very Important Faces: Yet Another Character Annotation Tool

Abstract

This short paper describes the ongoing project of creating yet another character annotation tool, the Very Important Faces (V.I.F.) tool. Although the idea of character annotation is really not a new subject, off-the-shelf software annotation tools have proved to be designed for contexts where the assumptions are not the same as in the case of historic photographic catalogs. Thus, the adoption of such tools has shown, in practice, to be below the expectations. The most evident limitation of the majority of the available photo annotation tools is that they do not process information present in captions and texts produced by experts that describe the contents of the photographic collections. Our dataset is constituted of a contemporary historic character photographic collection, with informative captions, available for public access . The design proposal of the V.I.F. tool is to help the experts responsible for collection organization to migrate the information, documented in the texts associated to the images, to W3C metadata standards. The V.I.F Tool implements face detection algorithms. It also detects proper names in previously inserted captions to help the user (expert) match names and faces in order to produce a photo annotation compatible with semantic web principles.