Rappler Uses Ontotext GraphDB to Power The First Philippine Politics Knowledge Graph

In 2022, Rappler used Ontotext technology to bring a new approach to publishing and managing their content that supported their efforts towards a more transparent and informed election process

  • Providing more complex information to their audience
  • Working with both historical and future election data for deeper analytics of events
  • Tracking the performance of government leaders
  • Finding hidden relationships between disinformation topics, actors, and projects, and discovering trends and dependencies

The Goal

The leading Philippine digital-native newsgroup Rappler, founded by Nobel Prize winner Maria Ressa, wanted to be able to provide more complex information to their audience and perform deeper analysis of political and economic events.

In the short term, the media company aimed to analyze and present data about the 2022 National Elections while laying the foundation for working with both historical and future political data. They wanted to create a central body of knowledge for each party, candidate, and location. To do that, they needed to model the Philippine political and electoral system – a highly complex hierarchical system with a large number of classes and relationships.

The Challenge

There were various challenges to the task such as Rappler’s small newsroom team, the sparse data for local aspirants coming from readily available data sources, and complexities in hierarchical relationships and demographic information across administrative areas.

On top of that, the media company needed to generate nearly 50,000 pages that would be served to a public front-end. These had to reflect an understanding of which contests (local and national) applied to each locale, and which candidates were running in each contest.

In addition, they needed to augment their editorial team by generating natural language profiles for 46,165 local candidates to preserve the factual integrity of Rappler’s research.

The Solution: An Ontology of Philippine Politics

In 2022, Rappler chose Ontotext GraphDB to store and publish their data. This enabled them to construct the first Ontology that captures all complex relationships of the Philippine political and electoral system. The media company used data gathered by Rappler’s Research and Data teams and combined GPT-3 Curie with their own Natural Language Generation system. This ensured a variety of content while mitigating the risk of generative AI hallucination to construct data-driven profiles.

High-level architecture and data flow

 

Ahead of the 2022 elections, Rappler’s Philippine Politics Knowledge Graph provided data for:

  • 83 national candidates with rich profiles combining data visualization and editorial expertise

  • 46,165 local candidates with data-driven natural language profiles

  • 1,732 locales (regions, provinces, cities, municipalities) with natural language profiles

The following example is taken from real data uploaded during Rappler’s Data Ingestion Workshop on July 21, 2022. Thanks to the Politics Ontology and GraphDB’s inferencing capabilities, they are now able to identify politicians associated with areas beyond their immediate constituency according to the Philippine administrative hierarchy.

Business Benefits

  • Providing more complex information to their audience
  • Working with both historical and future election data for deeper analytics of events
  • Tracking the performance of administration
  • Finding hidden relationships between disinformation topics, actors, and projects, and discovering trends and dependencies

Why Choose Ontotext

One thing that we realized from our research on disinformation and the information ecosystem is that there is a critical need to contextualize and link different pieces of information and stories in order to surface meaning and nuance. Ontotext is helping us achieve this by making it easier for us to build our own Knowledge Graph that precisely aims to do this.

We have already used this for deploying our election executions where we linked stories we produced in the past with other data sets in order to create highly engaging profiles of candidates. Taking off from the election coverage, we are excited to use this to programmatically track the performance of each administration, to surface relationships between disinformation narratives and actors, and other key projects of Rappler” – Gemma Bagayaua-Mendoza, Head of Digital Services and Lead Researcher for Disinformation and Platforms at Rappler Inc.

Do you think this case resembles your particular needs?

New call-to-action

Contact Us Now