Atomic has been originally developed in the context of Towards a corpus-based typology of clause linkage - An analytical framework and case studies on non-local dependencies (LinkType), a DFG-funded linguistic research project that is jointly carried out by the universities of Zurich and Jena. The project aims at statistically modeling linguistic variation in the area of complex sentences. It is based on the investigation of corpora from genealogically and geographically diverse languages that have been richly annotated.
Atomic is built as an open source annotation software for multi-layer deep linguistic annotation of text corpora. It is based on the Eclipse RCP, a modular rich client platform implemented in Java, and hence extensible via plugins.
Atomic works on a concrete implementation of the generic graph-based meta-model Salt, and embeds its complementary conversion framework Pepper, allowing for n : m mapping between data formats.
In the context of the LinkType project, Atomic has been developed as an architectural prototype, i.e., as a test bed for an extensible, sustainable annotation platform. Hence, all current functionality must be considered experimental, i.e., unstable and due to change.
Currently, Atomic is being developed towards an initial stable release.
To track progress, have a look at the development branch on GitHub. We aim for a transparent development process via issues and the Features project on the GitHub site. Feel free to file an issue if you want to contribute.