Project

Network of searchable leaks (new/s/leak) is a research project producing a software prototype that allows to quickly and intuitively explore large amounts of textual data. The tool will support journalists working with datasets like the Afghan War Diary, the U.S. diplomatic cables leak (both distributed by Wikileaks in 2010), or the hacking of The Hacking Team (2014). The goal is to provide a quick access to important entities (people, organizations, places) and their relationships, and how their usage changes over time.

Collaboration

The new/s/leak project is a collaboration between Hamburg University (Department of language technology), Darmstadt University (department of information visualization) and Der Spiegel. In 2016 and 2018, the project was funded by the Volkswagenstiftung as a part of the initiative “Science and Data Journalism”.

Recent News

New/s/leak 2.0: Second project phase finished

We are happy to announce that the second version of the tool new/s/leak has been completed. New/s/leak supports journalists in the evaluation of very large volumes of documents, such as those that repeatedly emerge from leaks of internal company or government data. The approach of the open-source software is the automatic identification of proper names as well as their visualization and filtering along relationship networks in the documents. In this way, one can quickly gain insight into otherwise unmanageable amounts of data and find starting points for journalistic reporting.
read more

New/s/leak at WissensWerte 2018

We present new/s/leak at a panel discussion at WissensWerte 2018, Germany’s most important dialogue forum for science journalists. On November 20, 2018 together with panelists from journalism, IT startups, and other universities, we will discuss how artificial intelligence contributes to journalistic work. In case of new/s/leak, we employ machine learning to automatically extract relevant information such as named entities and keywords from texts. This enables us to create interactive comprehensive visualizations of large text collections which contribute to a fast exploration for investigative purposes.
read more

More

Paper accepted at EMNLP 2018 (Brussels, Belgium)

Paper accepted at SocInfo 2018 conference in St. Petersburg

Newsleak 2.0 pre-release software demo

Presentation at #EIJC18 & Dataharvest conference

All News