ELAN

ELAN is being developed at the Max Planck Institute for Psycholinguistics in The Language Archive (TLA). It is written in the programming language Java; the sources are available for both non-commercial and commercial purposes. With ELAN, a user can create an unlimited number of annotations to primary audio and video data. An annotation can be, for example, a sentence, word or morpheme, gloss, a comment, a translation, a gesture type, or a tag or description of any other type of feature observed in the media.

ELAN offers a layered or tier-based annotation system. These tiers can be hierarchically connected. On independent tiers, annotations are arranged synchronously or time-aligned to the media – the user determines or changes the start and end point of each annotation. On dependent tiers, further down in the hierarchy, annotations refer to other existing annotations, whereby they can either use the same start and end times (one-to-one relationship), or further subdivide the interval of an annotation on an tier higher up in the hierarchy (1:n relationship). For example, a tier can contain annotations with sentence-related transcriptions while one-to-one dependent tiers may contain translations of those sentences into other languages. Or a dependent 1:n tier contains an annotation for each word or morpheme of the tier at the higher level, while another includes a gloss for each word or morpheme. Numerous 1:n relationships are possible, e.g. depending on whether internal start and end times can be determined on the dependent tier. This flexibility within the tier relations allows to model the complex ontological relations of features that are annotated on different tiers. The hierarchical configurations can be specified for multiple speakers and can serve as a template for new annotation documents. Annotations can be queried by either a single or a multi-tier search, with or without regular expressions, within a single document or within a selection of annotation documents. The annotations are always created and stored in separate files, independent of the associated multimedia files (stand-off); the latter remain unchanged. The textual content of annotations is always in Unicode and the annotation files are stored in a special XML format, “EAF” (ELAN Annotation Format). When archived in The Language Archive, ELAN data can be displayed online in a web browser together with the associated media files.