Curated

In the second phase of CEDIFOR, we continue the associated projects and initiate curation projects that are carried out through direct support and participation by CEDIFOR using the expertise gained in the pilot projects.

Referatorium for Qualitative Research Software

Connection: DIPF | Leibniz Institute for Research and Information in Education

Financed by means of the funding institution: Jan 2018 – Dec 2019

Project staff

Cornelia-Flavia Veja, DIPF
Julian Hocker, DIPF
Other participants: Christoph Schindler (DIPF)

Description

The project deals with the visibility and reuse of software for qualitative research. The focus here is on the implementation of the FAIR principles, with a special focus on the aspect of findability. The Semantic MediaWiki platform is used, which enables a participatory approach both for the creation of the content and for the input masks. The Referatorium is supplemented with workflow descriptions to make the application easier and more transparent.

Resources

Relevant publications

 

Open Data for Educational Research

Connection: DIPF | Leibniz Institute for Research and Information in Education

Financed by means of the funding institution: Jan 2018 – Dec 2019

Project staff

Cornelia-Flavia Veja, DIPF
Julian Hocker, DIPF

Other participants: Christoph Schindler (DIPF)

Description

Open data is a growing field in Germany. Unfortunately, the quality of the data varies greatly depending on the state or city. This project would like to explore the potential and limits of the use of open data in educational research. For this purpose, data relevant for educational research is being collected, examined, and combined. Another goal of the project is the evaluation of the existing data and the potential for further research projects and applications.

Resources

Relevant publications

URLCoFi (Under-Resourced Language Content Finder)

Connection: Goethe University

Supported with funds from the central eLearning facility at Goethe University: studiumdigitale – Aug 2019

Project staff

Armin Hoenen (GU)

Cooperation partner

studiumdigitale (financing, supervision), various linguistic institutes

Description

The project is dedicated to finding Internet content in languages for which little electronic resources are available which in turn are difficult to find. The primary goal of the project is to develop a Flipped Classroom course. In addition, a software solution is also created that recommends particularly promising word groups and word combinations based on certain parameters of a user request.

Resources

eLearning course

Relevant publications

Motif Analysis Toolkit for Text-Based Graphs

Connection: TU Darmstadt

Project staff

Dr. Thomas Arnold, UKP Lab, TU Darmstadt

Cooperation partner

Pascal Jürgens, Institute for Journalism, Johannes Gutenberg University, Mainz

Description

Motif analysis counts the number of recurring patterns (also called motifs) in a graph and connects these statistical numbers with the intrinsic semantics of the graph. The composition of the different motifs in a text results in a motif signature. In numerous experiments, correlations between these signatures and the properties of the underlying texts have been demonstrated, including various facets of text quality.
In order to make these analyses accessible and easy to use, the individual steps from unformatted text to statistical analysis of the motifs were combined in a motif analysis toolkit. This toolkit offers an intuitive user interface. Furthermore, all intermediate steps can be displayed, including the graph structure for visualization. The format of the input texts, the graph representation and also the motif search are modular and can be easily expanded.
In this project, the functionality of the motif analysis toolkit is evaluated on various data sets of political speeches and discussions. The connection between motives and different characteristics of the speakers is examined.

Resources

Relevant publications

Knowledge Base Supported Annotation with INCEpTION

Connection: TU Darmstadt

Project staff

Dr. Thomas Arnold, UKP Lab, TU Darmstadt

Cooperation partner

Markus Schäfer, Communication Studies, Johannes Gutenberg University, Mainz

Description

The need for semantically annotated text corpora is greater than ever. “Smart”, flexible, and reusable annotation tools are required to get annotated data in the most cost-effective way. With INCEpTION we want to build an annotation framework that integrates all related tasks into a common web-based platform.
The basic functionality is to perform annotations at the text level. In order to support this task sensibly, INCEpTION offers possibilities to connect and expand knowledge databases. The knowledge base enables the creation of cross-document links, for example to link people, places or other entities in the text with the entries of a selected database. We specifically target structured knowledge – i.e. not just flat or hierarchical annotations, but entity classes and entities that have properties and can also be linked.

Resources

Relevant publications

Multitask Learning Framework

Connection: TU Darmstadt

Project staff

Erik-Lân Do Dinh, UKP Lab, TU Darmstadt

Cooperation partner

Description

In this project, a multitask learning approach from a specific application into a generally applicable framework is implemented and thus made available to a broad user base. Multitask Learning pursues the idea of training a neural network for several tasks at the same time. A common part of the network is simultaneously optimized for all tasks to be solved, which creates synergy effects. In this way, significant improvements could be achieved in numerous experiments. This project is based on a multitask setup, in which several tasks for metaphor recognition in German and English have been combined. Within the framework of the Curation Project, this is generalized into a flexible framework and applied to new problems.

Resources

Relevant publications