10. SWISSUbase – Taking Quality to the Next Level
Imagine a researcher who needs to deposit research data. She wants easy access to a list of widely used, high-quality, and well-documented repositories so that she can choose which one is the best for her needs.
An overview of available repositories, including their maturity level, is essential to assess the quality of the open research data infrastructure. Additionally, it would act like a central hub to provide guidance on standards. Let’s hack on this!
11. Research Data Connectome
Scientific data is stored in many repositories which are not necessarily interoperable. Integrating metadata into a Linked Data Knowledge Graph solves this, which is a goal of the Connectome project.
We are looking for diverse teams of researchers and developers with Text-Data-Mining or Natural-Language-Processing expertise. The aim is to implement features for automatic entity-extraction from abstracts and their integration into the Connectome’s Linked Data Pipeline.
12. App to import all metadata in a reference management software
Research data are often not correctly cited in scientific articles and often omitted in the bibliographic section, despite the fact that repositories suggest a data citation and attribute a PID. The idea is to build an application that allows the import in a reference management software (e.g. Zotero, EndNote) of all the metadata needed to build the data citation in different bibliographic standards so that the researchers could save and use data citations as they would do for any other types of document (article, chapter, etc.). This would ease in-text referencing and the correct data citation in the bibliographic section.
13. Machine learning to cluster images based on their visual content
Image databases are mostly text databases with attached images. However, there are many image collections that have not yet been cataloged and are as such not searchable. Providing descriptive metadata to images is cumbersome and tedious. It would be extremely helpful if the images can be pre-ordered, that is clustered, based on image content. E.g providing clusters of similar images regarding the visual appearance would be great. Let’s explore how machine learning can help to cluster images according to shape, color, forms, etc. The DaSCH will provide some large collections of digitized photographic images for this purpose.