Ms. Plank, can you tell us a bit more about yourself and your assignment at TIB?
I studied Applied Information Science and Media Studies at the University of Hildesheim, specialising in information retrieval and usability. After graduating, I was responsible for a variety of eLearning projects at the University of Hildesheim as a research assistant.
I came to TIB in July 2007. I first worked as an information literacy expert and became Head of the Competence Centre for Non-Textual Materials in 2010. I first came into contact with multimedia retrieval as a student. In this context, I was particularly fascinated by automatic image recognition processes – i.e. technologies enabling large data volumes to be found using just image information such as colours, contours and textures.
I recall that, back then, research was faced with the great challenge of automatically differentiating between images of the Love Parade in Berlin and images of carnival celebrations in Cologne: two completely different concepts that humans can differentiate between at a glance – but that machines were unable to resolvable, unless Berlin’s Victory Column or Cologne Cathedral happened to be in the picture.
KNM was established in 2010. In a nutshell: what are you working on there at the moment?
The Competence Centre for Non-Textual Materials – KNM for short – has been in existence for four years. At present, the team comprises nine members of staff with completely different core competencies: IT developers, multimedia retrieval and ontology specialists, media documentalists, information managers and in-house lawyers belong to the team.
We work hand in hand to achieve the fundamental improvement of conditions surrounding the access and use of audiovisual material, 3D objects and research data. We would like to provide active support to our users in their scientific work process, making it as easy for them to publish, locate and sustainably provide non-textual material as it is for them to procure textual documents. We provide support to other knowledge institutes and knowledge providers in all matters concerning non-textual material, providing services, tools and infrastructures.
All of the projects we implement here have to do with visual search, automatic indexing methods and visualisation. We put into practice approaches that are elaborated in research projects. In PROBADO, we achieved the semi-automatic indexing of 3D models and a search using room connectivity graphs in 3D models. In this case, the project partners were three Computer Science chairs at the universities of Bonn, Graz and Darmstadt, and Fraunhofer Institute for Computer Graphics (Fraunhofer IGD).
In the VisInfo project we joined forces with the Interactive Graphics Systems Group (GRIS) at Technische Universität Darmstadt and Fraunhofer IGD to implement a similarity-based visual search in research data.
We have just completed another project – the TIB|AV Portal – a portal for audiovisual media that combines state-of-the-art multimedia retrieval methods with semantic analysis. The topic of usability is of great personal interest to me. In order to ensure the optimum usability of portals, we always ensure that users are involved in our projects from the very beginning.
The TIB|AV Portal has been online for a few days now. What is the initial feedback like?
What is special about the portal is that you can search not only for entire videos, but even within actual videos. If, for example, a user searches for the term ‘wind power station’, not only the metadata, but also the language, text overlays and image information will be searched. The detailed view of a video then shows exactly in which segment the speaker mentions wind power stations, enabling the user to jump straight to that part of the video.
Feedback about the TIB|AV Portal has been fantastic. A number of users have sent us suggestions and ideas about how to further develop the portal via the feedback button integrated in the portal. Several users, for example, would like to be able to filter results by Creative Commons licenses; others would like to see an OAI interface – OAI stands for Open Archives Initiative – to enable them to make further use of the data. At the moment we are collecting all of the suggestions, which we will then prioritise.
The TIB|AV Portal currently contains around 2,000 science and technology videos, as well as some 1,000 film credits with external links to other portals. New scientific films such as computer animations, project documentation and recordings of lectures and conferences are being added to the collection every day. We last imported films from Khan Academy and a large package from the German Aerospace Center (DLR).
Users can also upload their own films onto our portal. For large film collections, however, it makes sense to request FTP access data from us. Films are tested for quality, indexed according to international standards and given a digital object identifier (DOI). DOI names ensure that films can be clearly cited, even at the segment level. This way, individual video sequences can be cited as easily as the page of a book.
The Competence Centre for Non-Textual Materials (KNM) belongs to the area of “Development” at TIB, and is therefore part of the “Department of Research and Development”.