Open Justice

NPA Accountability:


the media and citizens through data

Open Justice sought to tackle NPA accountability through several data and tech experiments. The project aimed to make new forms of NPA information, such as court data, open and available for both journalists' analysis and inform citizens of important NPA activities.

Share this project

About this project

The National Prosecuting Authority (NPA) is responsible for instituting criminal proceedings on behalf of the State and combating corruption and economic malfeasance. In a report to Parliament, the NPA identified its priorities as restoring its credibility and strengthening communication. Rebuilding the NPA as an independent, professional, accountable, and credible organisation requires empowering the media to enhance credibility and accountability.

OpenUp aimed to support accountability through the following initial outcomes:

  • New forms of NPA information, court and data are made open for analysis; 
  • Citizens are informed of NPA activities in a contextualised and relevant manner through the provision of content for improved accountability;
  • The capacity of a journalistic cohort is improved to both enrich the information gathered, and report it in a way which enhances the accountability of the NPA while improving the quality of information the public receives;
  • The NPA is activated to provide additional direct access to documentation that facilitates better reporting by our journalistic cohort for improved citizen access to NPA-related, enriched content.


Research Report:

The NPA has become an important civic focal point, following the Zondo recommendations. This has proved both an opportunity, and a challenge. In January 2022, OpenUp participated in direct engagements with the NPA’s Anti-Corruption Task Team as part of civil society engagements. In that engagement, we specifically raised issues of data access and data collection. Whilst this was a public expression, we haven’t observed any dramatic change in data provision over the project period. The main accountability challenges for the NPA tend to exist vertically between the NPA and Parliament rather than with members of civil society, and the public.

In response to challenges in accessing NPA-related materials (and heavily influenced by our concerted efforts in accessing court records in our previous Open Justice project - you can read about some of experiences in that process here: we decided that there were two distinct knowledge needs for this objective:

  1. Exploring the different dynamics of accountability to better understand what form data accountability might take; and
  2. Explore existing alternative forms of NPA data accountability. 

To do this, we analysed the NPA’s Twitter feed and analysed this data whilst unpacking accountability in a groundbreaking report called “Tweeting Accountability”, which is currently in design. The research mapped data across the following dimensions of accountability:

  • Upward Accountees — those to whom the NPA is upwardly accountable (e.g., Parliament).
  • Downward Accountees — those to whom the NPA is downwardly accountable (e.g., the public).
  • External Motivation — an imposed motivation for the NPA to be accountable (e.g., the Constitution).
  • Internal Motivation — a felt motivation for the NPA to be accountable (e.g., the NPA’s internal accountability chain, policies).
  • Formal Mechanisms — instrumental mechanisms for providing accountability (e.g., Reports).
  • Informal Mechanisms — expressive mechanisms for providing accountability (e.g., Media)
  • Long-term Time Frames — strategic accountability, usually to Upward Accountees (e.g.Annual Reports).
  • Short-term Time Frames — functional accountability, usually to Downward Accountees (e.g., Tweeting updates about a case with strong public interest).

In so doing, the research demonstrated that, via its Twitter account, the NPA does reference (to varying degrees) Upward and Downward Accountees and the formal mechanisms for providing them with accountability. That said, Twitter is, by its very nature, an informal mechanism for providing accountability and can never replace the formal mechanisms already in place. For this reason, Twitter can only be used to provide public accountability concerning the NPA.

It allows the NPA to provide the public with (1) an account of the ways in which its prosecutorial responsibilities are being met, and (2) an account of its corrective actions (when such actions are required). In this regard, the NPA shares updates of its prosecutorial responsibilities via ‘media’ or ‘press’ releases related to cases it is involved in (roughly 67% of all the tweets analysed).

This research also provides a framework that OpenUp can use to explore different types of NPA accountability as more data streams emerge. 


One of the most significant developments under this grant has been the ideation and creation of “Dexi”, which was a result of our hard pivot from the Aleph instance originally refined for our Open Justice project.

Often we have ample material but not enough time or resources to digest it. This is where Dexi shows value. Dexi is a Dynamic Entity Extractor and Indexer. Using industrial-strength natural language processing tools, Dexi can recognise and extract entities like people, organisations, countries, cities, nationalities, political parties, laws, etc. 

But it doesn't just recognise entities. Give Dexi a document, and it will collect, classify and index the entities it finds. This means even long documents can quickly become useful and navigable resources. Dexi uses Spacy for natural language processing and can extract and recognise entities such as people, organisations, countries, cities, nationalities, political parties, laws and more from documents. Dexi can then build an index from single or multiple documents. 

A Dexi project can have multiple related documents and extractions. Dexi can track entities accross all the documents in a project
Extracted Entities are counted, indexed and colour-coded in the document view according to entity schema
Dexi's modular approach to entity extraction and text analysis, means it can do various visualisation and computations on entity sets and text, including basic word-clouds.

Where Dexi surpasses the capabilities of Aleph is in its choice of natural language processing models. Dexi was built around the requirement of being able to interchange different language models for different use cases. Different models can focus on different areas, including science, law and sentiment analysis.  

Pocket Reporter:

OpenUp developed the Open Justice Pocket Reporter tool for economic crime reporting to guide and support journalists in collecting information for investigations. Pocket Reporter can be adapted and used as a court reporting tool for journalists and members of the public following NPA related cases. You can view the adaptation here: 

Pocket Reporter leveraged existing South African National Editors’ Forum ( SANEF) materials to develop this specific corruption-related court reporting subsection. This was shared with both SANEF and the Association of Independent Publishers.  

We consider Pocket Reporter an “App-In-Waiting” for the right political and journalistic moment. The investment in the update should see impact in the future as OpenUp begins reinvigorating its media activities, buttressed by our other work under another funded project called the African Data Hub.

Project details

OSF South Africa
Project start:
January 2022
Project status:
No items found.

Project video