The CAMOMILE Collaborative Annotation Platform for Multi-modal, Multi-lingual and Multi-media Documents

The CAMOMILE Collaborative Annotation Platform for Multi-modal, Multi-lingual and Multi-media Documents Johann Poignant1, Mateusz Budnik2, Herv´e Bredin1, Claude Barras1, Mickael Stefas3, Pierrick Bruneau3, Gilles Adda1,6, Laurent Besacier2, Hazim Ekenel4, Gil Francopoulo6, Javier Hernando5, Joseph Mariani1,6, Ramon Morros5, Georges Qu´enot2, Sophie Rosset1, Thomas Tamisier3 1. LIMSI, CNRS, Univ. Paris-Sud, Universit´e Paris-Saclay, F-91405 Orsay, firstname.lastname@limsi.fr 2. LIG, Univ. Grenoble Alpes, Grenoble, France, firstname.lastname@imag.fr 3. LIST, Esch-sur-Alzette, Luxembourg, firstname.lastname@list.lu 4. IT ¨U, Istanbul, Turkey, lastname@itu.edu.tr 5. UPC, Barcelona, Spain, firstname.lastname@upc.edu 6. IMMI-CNRS, Orsay, France Abstract In this paper, we describe the organization and the implementation of the CAMOMILE collaborative annotation framework for multimodal, multimedia, multilingual (3M) data. Given the versatile nature of the analysis which can be performed on 3M data, the structure of the server was kept intentionally simple in order to preserve its genericity, relying on standard Web technologies. Layers of annotations, defined as data associated to a media fragment from the corpus, are stored in a database and can be managed through standard interfaces with authentication. Interfaces tailored specifically to the needed task can then be developed in an agile way, relying on simple but reliable services for the management of the centralized annotations. We then present our implementation of an active learning scenario for person annotation in video, relying on the CAMOMILE server; during a dry run experiment, the manual annotation of 716 speech segments was thus propagated to 3504 labeled tracks. The code of the CAMOMILE framework is distributed in open source. Keywords: Annotation tool, collaborative annotation, multimedia, active learning, person annotation. 1. Introduction Human activity is generating growing volumes of hetero- geneous data, available in particular via the Web. Multi- modal, multimedia, multilingual (3M) data can be collected and explored to gain new insights in social sciences, lin- guistics, economics, behavioural studies as well as artifi- cial intelligence and computer sciences. But to be analyzed through statistical-based machine learning methods, these data should be available in very large amounts and anno- tated. Annotating data is costly as it involves manual work, and in this regard 3M data, for which we need to annotate different modalities with different levels of abstraction is especially costly. Current annotation frameworks often involve a local man- ual annotation, e.g. LDC annotation tools (Maeda and Strassel, 2004), Anvil (Kipp, 2001), Viper1 (Doermann and Mihalcik, 2000) sometimes supported by automatic pro- cessing as proposed by ELAN (Auer et al., 2010). In this case, dealing with multiple annotators and different ver- sions of the annotation files quickly becomes unfeasible. Browser-based annotations interfaces, linked to a server for storing the annotations, can provide a solution to this prob- lem, all the more so as browser performance and multime- dia support dramatically improved in the recent years; this is a direction taken e.g. by LDC with its WebAnn initiative or by the PubAnnotation project2 which is a repository of 1http://viper-toolkit/sourceforge.net/ 2http://www.pubannotation.org text annotations and associated services in the domain of life science literature. In the context of the CHIST-ERA CAMOMILE project (Collaborative Annotation of multi-MOdal, multI-Lingual and …

Content truncated to 500 words