AVATecH (Advancing Video/Audio Technology in Humanities Research)

From all the proposals presented to the innovation funds in 2008 that want to stimulate collaboration between Max Planck and Fraunhofer Societies, the AVATecH (Advancing Video/Audio Technology in Humanities Research) project received the highest rating. In a collaboration between the MPI for Psycholinguistics and the MPI for Human Ethnology on the one hand and the HHI and IAIS Fraunhofer institutes on the other hand want to close a gap that that is increasingly obvious. Despite all efforts to improve digital speech and image recognition during the last decades there is effectively no progress for helping the field worker when analyzing his material. The reason for this is that all assumptions for mainstream recognition technology are not fulfilled: we are working with non-standard languages where we don't know the language model and where often the recordings are very noisy.

Therefore it was proposed to start from scratch again and to build first a number of simple detectors for segmental and supra-segmental speech and for gesture and motion patterns that can be integrated in an interactive framework such as ELAN. These detectors create annotations and with the help of a flexible and powerful search engine these patterns can be analyzed. In addition such detectors can be combined to new ones and more complicated can be created later. The two Fraunhofer institutes are experts in the areas of speech and image recognition and have already a number of simple and complex detectors that can be integrated. The MPI in Nijmegen has with ELAN already a powerful and flexible engine that can integrate detector components and that allows to search on complex patterns.

The hope is that soon the researchers can select such detectors from a library and test them to facilitate the tedious annotation work.