Artificial Intelligence (AI) has seen major advances in the last 12 months. Systems like ChatGPT, MusicLM or Midjourney have demonstrated near-magical capabilities in the generation of written and audio-visual content. However, these are not monolithic systems and are constructed of various elements, many of which are open-source or have open-source substitutes.
Without getting too technical, computational biology and comparative linguistics have greatly advanced in recent years and present a great opportunity for the media research space.
With the goal to deliver optimal earned media outcomes for our clients, here is how AI is shaping our research initiatives.
Natural Language Processing
In March of this year, Medianet switched on Australia’s first AI-assisted New Author Signal research system. In just one day it increased our media database updating productivity by 1203%.
To achieve this we work within a subset of Artificial Intelligence known as Natural Language Processing, which is concerned with the computational analysis of human writing and speech.
Our New Author Signal system utilises a modified neural network architecture, which has been tailored to the workflows of Medianet's Research team.
This new system works by taking publicly available metadata (which we refer to as a signal) through a series of algorithms in order to determine when journalists have changed jobs, entered the industry, changed their name, or various other pieces of information that are highly relevant to our clients seeking to maximise their earned media presence.
We then integrate these signals with our own task ticketing system in order to streamline the process into a single pipeline. Any time the algorithm determines that someone's records in our media contact database are incorrect, it generates a ticket for one of our Researchers to reach out to the media contact to either confirm or reject the suggestion.
By incorporating a human-AI hybrid approach, we find we get the best of both, letting automated systems grind through large magnitudes of data while letting human operators verify the data and make that personal connection.
["Neural Network: basic scheme" by fdecomite is licensed under CC BY 2.0]
Network Visualisation
The study of genetic biology may not sound related to the study of the media. However, when you break both areas down to their most basic elements, there is considerable overlap.
Both are concerned with the study of relationships through the transfer of information between nodes. These relationships form clusters, edges, and ultimately networks of information transfer. In the same way that various proteins may interact within a cell, various stories interact within a media landscape.
Through extensive research, we found many of these similarities in the systems and approaches from computational network biology utilised by various software platforms such as Cytoscape.
Our goal here was to better understand the connections and interactions within a media landscape, so our first application into network visualisation was through a deep dive into the New Zealand media landscape.
To kick this off we started by inputting our journalist and outlet data into Cytoscape, and then after some testing settled on an algorithm called the Edge Weighted Spring Loaded Layout, with some modifications. This approach gave us the opportunity to accurately represent the relationships between outlets and journalists in a visual way that was easy to understand.
As an Australian-based company that services both sides of the Tasman as well as Pacific Nations, it is important that we maintain all our records at a consistently high standard, regardless of where those contacts and outlets are based. Through this approach, we were able to identify very interesting aspects of the New Zealand media landscape that a standard analysis would never have picked up and then used that to direct our Researchers to update the relevant records.
["New Zealand Media Landscape Network Visualisation" created via Cytoscape from Medianet data]
Just the beginning
Natural Language Processing and Network Visualisation are only just scratching the surface of what is possible with AI. With new technologies continuing to change the media landscape in ways that are impossible to predict, we continue to adapt and optimise our research approach to deliver the most accurate and reliable data possible.
🤖🤖🤖
Share
Top