Paper accepted at SocInfo 2018 conference in St. Petersburg

Newsleak will be presented at the Social Informatics conference 2018 which takes place from 25-28th of September in St. Petersburg, Russia. A preprint of the conference paper can be found here.

Abstract: Investigative journalism in recent years is confronted with two major challenges: 1) vast amounts of unstructured data originating from large text collections such as leaks or answers to Freedom of Information requests, and 2) multi-lingual data due to intensified global cooperation and communication in politics, business and civil society. Faced with these challenges, journalists are increasingly cooperating in international networks. To support such collaborations, we present the new version of new/s/leak 2.0, our open-source software for content-based searching of leaks. It includes three novel main features: 1) automatic language detection and language-dependent information extraction for 40 languages, 2) entity and keyword visualization for efficient exploration, and 3) decentral deployment for analysis of confidential data from various formats. We illustrate the new analysis capabilities with an exemplary case study.

Newsleak 2.0 pre-release software demo

Since the first version of Newsleak, a lot has been improved behind the scenes as well as in the front-end of the software. We want to encourage journalists, to try out a pre-release of Newsleak 2.0 on their own. For this, we provide a software demonstration. This demo is populated with ca. 26,500 documents collected from Wikipedia in four languages (English, German, Hungarian and Spanish) and mostly centered on the topic of World War II. The idea behind this demo is to show you the analysis capabilities to quickly explore a large, multilingual collection.

For lazy clickers, we provide a Youtube video where you can follow a proceeding of an exploratory analysis and filtering process drilling down to some details of inner-Chinese political tensions during WW2.

Presentation at #EIJC18 & Dataharvest conference

This Saturday, we present new/s/leak at the European investigative journalism conference (EIJC). Here you can find the slides of our presentation about “Information Extraction and Visualisation for Investigative Journalism”.

If you are interested to try new/s/leak with your own data, visit the Github page containing the Docker setup of our application.

In June, we will publish a detailed blog post on how to setup Hoover and Newsleak to analyze collections on your own machines.

Dataharvest Conference #EIJC18

From Thursday 24 to Sunday 27 May 2018, the EIJC 2018 conference (European Investigative Journalism Conference) will take place in Michelen (Belgium). We as newsleak project will participate and discuss requirements and needs of our targeted user group. All about the conference you can find out on this website:

Funding extension

We are happy to announce that the new/s/leak project receives some additional funding from the Volkswagen Stiftung. Until summer 2018, new/s/leak will be extended and refactored to achieve the following goals:

  • easy deployment for own usage
  • comprehensive and detailed documentation
  • improved user interface
  • improved information extraction (better keyterm extraction, named entity recognition, support of user dictionaries)
  • support for multiple languages (among others english, german, spanish, french, arabic, chinese)

Follow the updates on this blog to see how far we got ūüôā


new/s/leak demo @ SPIEGEL

Now that we’re in the middle of new/s/leak’s home stretch, we had a final demo at SPIEGEL in Hamburg. After some exciting and productive development sprints, we¬†proudly introduced the software to journalists, documentarists and software developers, who gave us the best feedback by playing around with the tool and becoming absorbed in using it. Some evidence:

We also collected some more systematic feedback, which helped us prioritizing the remaining tasks. Thanks to everyone who came along, played and gave feedback – we had a blast at the meeting, and we learned a lot!

If you also want to see what changed  in new/s/leak since we have shown it to an academic audience at ACL: here is the link to the demo (please use the Chrome Browser!)

For a quick introduction, you can also watch a video (from our academic publication @ VIP):

During the upcoming weeks until christmas, we’ll add some more requested features, fix some bugs, and create an easy-to-deliver software package. Stay tuned for a deployable version!

new/s/leak @ VIP

Last week, new/s/leak had its academic debut in the visualization science community at the Visualization in Practice Workshop, co-located with the IEEE VIS 2016 conference.

Here is the paper¬†documenting the software with a focus on visualization. Needless to say that it’s always fun to present new/s/leak and get more feedback:

Kathrin presenting new/s/leak

Kathrin presenting new/s/leak

Thanks to everyone who came and visited us!


Paper accepted @ VIS 2016

Our Paper “new\s\leak — A Tool for Visual Exploration of Large Text Document Collections in the Journalistic Domain” has been accepted for presentation at the poster session of the Visualization in Practice Workshop, which is part of the IEEE VIS 2016 conference. The workshop will take place in¬†Baltimore Maryland, USA on October 24-25.

VIS is one of the most important conferences in visualization science. new/s/leak fits perfectly in this year’s VIP workshop, the focus of which is design, development, distribution, and application of open source¬†visualization and visual analytics software.

Meet us at the demo session in Baltimore!

new/s/leak @ ACL 2016

Last week, we presented new/s/leak for the first time in public: we had our demo session at the annual meeting of the Association for Computational Linguistics, which was held in Berlin this year.
If you haven’t had the chance to attend (or you had and want to have some references now):

  • Here’s the paper documenting our software (you’ll find a large part of the information from the paper in this blog, too)
  • Here’s our poster (in PDF format)

And, of course, we took some pictures, testifying how much fun we had (click for larger versions):

Alex walking through the new/s/leak poster (with Seid listening)

Alex walking through the new/s/leak poster (with Seid listening)

Seid explaining new/s/leak

Seid explaining new/s/leak

Chris, Heiner and Seid busy discussing new/s/leak

Chris, Heine, Seid and Alex discussing new/s/leak with different people

Seid and many curious new/s/leak fans

Seid and many curious new/s/leak fans

The crew busy explaining, demonstrating and (secretly) playing with new/s/leak

The crew busy explaining, demonstrating and (secretly) playing with new/s/leak

Thanks to everyone who stopped by, especially for the great suggestions for improvements! Be sure that we’re working on that while you’re reading this post. If you have any more ideas, application visions or simply want to debate information extraction and / or journalism – please get in touch!

The Science behind new/s/leak II: Interactive Visualization

We already explained the language-related data wrangling happening under new/s/leak’s¬†hood. For the success of new/s/leak, our second scientific field is the game changer: interactive¬†visualization. No matter how much accurate information we can produce – if we cannot present them to the user in an appealing way, the tool will fail its goals. So how exactly is visualization science influencing new/s/leak?

Your daily dose of visualization science

It might seem easy to create some kind of visualization (with Excel or even pen and paper) – however, there are lots of pitfalls that you need to avoid to create good visualizations. You might take some of them for granted because you encounter them everyday when browsing the web – but they’d be painfully missed otherwise. Two¬†examples which you can find in many applications and websites (and which we of course also consider for new/s/leak):

  • Animation speed: there is a certain animation speed that is pleasant to look at, and it cannot be much faster or slower if we want to convey information. While it’s intuitively clear that reaaallyyyy slooooowwww animations can be annoying (think of those endless progress bars…),¬†going to fast can overload you, even if there is no critical information involved. For evidence, take a minute look at those two guys working out (they are doing exactly the same movement).

    Now answer the question: would you pick the guy on the right as your office mate? Might become exhausting really soon (even if he’s really smart).
  • Colors: If you us colors for information visualization, you cannot just use any color set you find appealing. While there is scientific work on which color sets are good for which¬†information types, we also need to think about color blind people. On many web pages, the color sets are already designed to be suitable for different types of color perception. See how the new/s/leak logo would look for color blind people (created with Coblis):

Those are just two examples for a whole lot of guidelines that visualization scientists have developed and that we encounter in each good visualization. Of course new/s/leak has to follow all of those guidelines Рwhich becomes harder with more complex data, and on a scale. Which brings us to the next important point:


Accurate views on loads of complex data

This happens with too much information displayed at onceThe largest challenge (and thus the largest need) for visualization science and language technology alike comes with the huge amounts of data we have to handle. A leak can be anything from 100 Documents to 1TB (or more). This is not only a lot for research-based software, but also enough to break many commercial applications. So, this is where the action is.
Visualizing data for investigative purposes means that the software may not show anything untrue (not even shady, or too ambiguous). However, there is so much data to display (all the documents, their metadata, and all the entities we extracted) – we simply cannot display the whole truth in one screen, that would be a) impossible and b) completely unusable. Imagine a network that shows all the information it has – that quickly becomes a “hairball” like in the picture on the left.

Because new/s/leak should be intuitively accessible for users with different backgrounds and without much training, Usabilitywe need to find easy interfaces for giant piles of complex data.  This excludes e.g. some very powerful but rather complicated interfaces used for search in scientific environments.

One extreme way to tackle this is the way Google presents the internet to us, or rather the 60 trillion pages it has indexed: Initially, Google doesn’t display anything, but rather lets the user explore (see comic on the right). While we allow the user to explore the data on their own, new/s/leak’s main purpose is¬†to guide users through the data jungle and to provide a concise graphical summary of the core plots. In consequence, we display an initial graph as entry point that contains the most interesting entities and relations. It’s a scientific question of its own to find out what most interesting means, but according to our users, frequency is a good indicator here. We thus show initially the most frequent people, companies and places, and let the user explore from there. (If they wish, users can also pick the least frequent entities first – this might¬†foster serendipity.)


Knowing our users

Coming back to the usability comic – of course we have to design new/s/leak for its users. That’s what Apple and Google do, too: providing a simple one-fits-all screen as an entry point. While we have a smaller user group, we also have an application with more interaction possibilities. So the key to everything lies in knowing our users, in order to find the right mixture of the google window (with too little functionality) and an overloaded brick puzzle (which is ugly and unusable).

User studies are, in fact, an important pillar of visualization science. We already told you about early requirements management. It might sound trivial that we ask people what they like and dislike, and then we change the system accordingly – but this is no matter of course at all¬†(see e.g. some excuses why companies don’t do user research). One of the reasons for this is that it takes actually lots of experience to design a good user study, which assess all the information needed without influencing the user, and which allows to generate¬†meaningful hypotheses for interface changes. Further, it takes simply quite some¬†time to undertake such studies repeatedly. Fortunately, our visualization scientists¬†are experts for user studies, and they are dedicated to testing our system frequently and in a way that allows for objective, comprehensive evaluation.