The annotation of various sources is a crucial part of the research process in almost every field. While the classical form of computer annotation on electronic research papers is no surprise nowadays, one might not know that there also exists advanced tools for video annotation.
One such tool is ELAN (http://tla.mpi.nl/tools/tla-tools/elan/). The abstract Expanding and connecting the annotation tool ELAN relates how a network approach can be used to improve this tool. One of the main ideas for improving this open source software lies in the bridging of ELAN with an online multimedia lexical database, called LEXUS (http://tla.mpi.nl/tools/tla-tools/lexus/lexus-description/). This framework, named LEXAN (a mix of the two above names), appears very promising.
The benefits of using such a dual approach are plenty. Indeed, if one can easily enrich a given lexicon during the media annotation process, databases are bound togrow and diversify significantly. These various lexical databases could then be used by many other researchers from all over the world in their media annotation process.
A future challenge would be to improve the multi-document (semi-automatic) annotation functionality of ELAN, in order to face the growing quantity of available sources. While databases containing transcriptions and annotations of video corpora do already exist, there is hardly any agreement regarding the way to evaluate the quality of a given annotation database.
The growing need for automatic annotation tools is further emphasized in a second abstract: Automatic annotation of linguistic 2D and Kinect recordings with the Media Query Language for Elan.
This newly born language is based on the Kinect’s speech and movement recognition features. It creates a way to associate the recognition of a movement pattern (a person rising its left arm, for instance), with an automatic annotation of the recorded video file (for instance, from 3:10 to 3:15, annotate: “someone is raising its left arm”).
Further improvements were then made. For instance, the ability to recognize a given person, and to annotate the document only if this given person performs a prescribed action. The sensitivity of the pattern recognition can, of course, be fine tuned.
Even though this language is still a work-in-progress, many applications can be envisaged. One could, for instance, think about automatic textual transcription of sign language, or an analysis of the correlation between words or tone used in speech and the gestures accompanying them.
As a further note, the next generation Kinect promises a much improved accuracy in gesture and speech recognition. It would be interesting to see whether these new features can be taken advantage of.
Having exposed the need for automated annotation generation on large corpora, we should, however, not forget that other researchers might be more interested in extremely precise single-document annotation tools. The third abstract, The FAST-CAT: empowering cultural heritage annotations, elaborates on this radically opposed approach.
FAST-CAT is part of the CULTURA project (http://www.cultura-strep.eu/), whose purpose is to “bring historical texts to life”. It allows the annotation of images and text, as opposed to the movie annotation functionalities of ELAN. The differences, however, go beyond the type of media handled. FAST-CAT boasts the ability to finely tune the selection of content to annotate. Hence, for images, it is possible to select precisely a range of pixels to be enriched by an annotation.
Despite their philosophical difference, there is a strong connection between the recent developments of ELAN and FAST-CAT. They both aim to take advantage of a network connection in order to enhance the user experience. As a matter of fact, all of the annotations made using FAST-CAT are stored online. This not only increases portability of one’s work, but it also allows for collaborative annotation of a given document, which, in the case of complicated fragments of historical documents, is certainly a very welcome addition.
We have seen that various annotation tools are being actively developed with different goals in mind, although the direction of these enhancements is clearly directed towards the integration with new or existing online environments, and the automation of the annotation process for large corpora. It remains to be seen whether these improvements will be embraced by a community large enough in order to render them worthwhile.
Written by Antoine Imboden