Skip to main content
Date
2016-04-26

At this year’s Cebit, L3S demonstrates the results of the project ForgitIT organized by the team of our member Professor Wolfgang Nejdl and Dr. Claudia Niederée which provides the newest insights in Digital Forgetting themed “Humans Forget – and Computers?”

Nowadays we are facing a strong increase in the production of digital data, e.g. photos taken during events and everyday life. Although archived on cheap storage devices, data collections often end up as a kind of large dark archives, which are rarely accessed and enjoyed again due to their large sizes. Moreover, digital data is subject to a new and less obvious type of fragility, which leads to a form of random digital forgetting: over decades storage devices break down and formats and storage media become obsolete, making random parts of data collections inaccessible.

For dealing with this and similar situations of information overload for short-term and long-term information management, we took inspiration from the mechanisms of human forgetting and remembering. Actually, humans are very effective in focusing on the important things, while forgetting irrelevant details and filtering out redundancies. For this purpose we present managed forgetting, a mechanism of controlled digital forgetting based on multi-dimensional information value assessment. The idea is to offer an alternative to the dominant keep it all idea, which is prone to digital forgetting and dark archives problems.

As a showcase, we apply managed forgetting to support users in selecting important photos from their collections, to ensure that they remain accessible and enjoyable on the long run. We developed a novel photo selection method, which combines advanced multimedia analysis methods including deep learning with an expectation-driven approach. The main goal is meeting human expectations and preferences when selecting photos from a collection that are most important to them, for instance for revisiting or preservation purposes. The system uses machine learning based on observed selection behavior as a basis for identifying the most important photos from unseen collections. For increasing the applicability of the method in realistic settings, it does not impose any time consuming task to the user such as photo tagging and annotation with text.

For more information please have a look at: https://www.l3s.de/projects/forgetit

Contact: Dr. Claudia Niederée, niederee@L3S.de

At CeBIT 2015, L3S will present the newest insights from this area of Digital Forgetting at the research booth of the Federal State of Lower Saxony. If you like to visit the L3S-team at our booth in hall 6 A 18 at Cebit, please contact our project assistance at info@L3S.de; we are glad to assist you!