As expected, the News Angler project has generated new theoretical knowledge in several areas. A key contribution is a new understanding of and operational knowledge about news angles and how they can be formally represented and processed using AI-related techniques. Another contribution is a new understanding of and operational knowledge about how new AI techniques can be utilized in journalistic knowledge platforms. The project has also systematized and provided an overview of key theory and technology areas, such as: the use of knowledge graphs for news production and distribution; representation of named entities in knowledge graphs; and trustworthy use of AI in news production.
A significant technical outcome of the project is the News Hunter platform, made available as open-source for research and educational purposes [1]. Our News Hunter platform has continuously gathered news-related information from textual sources such as social media, commercial news aggregators, and open reference works. This information has been integrated and represented in network form, enabling easier, more precise analysis and sharing in new ways.
Furthermore, the project has delivered several openly accessible ontologies: Item Annotation Ontology, Event Annotation Ontology, and News Angle Ontology. Several examples of news angles have been analyzed and described in formal and conceptual detail in international publications.
We have thus achieved our central objectives: we have developed a big data-ready architecture for and a prototype of a journalistic knowledge platform (the News Hunter platform and the doctoral thesis); we have conducted pioneering work on journalistic angles and how they can be computationally represented and processed (multiple journal publications, including collaboration with The Open University); we have provided an example of how traditional symbolic and newer sub-symbolic techniques can be used together for this purpose (combined use of ontologies, knowledge graphs, and ML/DL-based language analysis techniques); and we have established new international, interdisciplinary, and industrial research collaborations (as described below).
Starting the project in 2018, it proved harder than anticipated to represent news-related texts in detailed knowledge graphs because the accuracy of relation extraction progressed slower than expected in the early years of the project period. Therefore, the current News Hunter platform does not support general automatic angle detection as initially envisioned. Such angle detection has come much closer over the past year through the availability of large language models like ChatGPT, which we are already investigating in our further work.
[1] Opdahl, A. L., Gallofré Ocaña, M., Tessem, B., Al-Moslmi, T. A. A., Holgernes, O. J., Asplem, S., Aamodt, O. K., Koren, U. W., Johannessen, M. S., & Åldstedt, S. (2022). The News Hunter Platform. Software, CC BY-NC-SA 4.0. https://git.app.uib.no/i2s-public/news-hunter-platform