We’re proud to say that this year at the workshop for Digital Humanities (NLP4DH), our industrial PhD student Anton Eklund presented a method for collecting human interpretation of different contexts with a new tool called Cluster Interpretation and Precision from Human Exploration (CIPHE) [1]. The tool presents a few articles as a news context, and then records human participants’ reactions to it. With that data, we can both calculate classifier accuracy, as well as record interesting characteristics of the context such as engagement and negative emotion.
Currently, we’ve undertaken news context analysis on the following characteristics: engagement, negative emotion and perceived impact on society. What we can see from preliminary results is that, generally, contexts linked with negative emotion also tend to have high engagement. The highest engagement was found in War & Conflicts and Hospital & Health, but also in lighter topics such as Food & Drink, Weather and household economy-related topics (though, interestingly, not the stock market!). The lowest engagement came from contexts like TV and Gossip/tabloids.
The “safest” (least controversial) contexts with the lowest negative emotion but okay engagement were Sports, Tech, Music, and Gardening. This provides crucial insights on how to best use differing contexts to your advantage. For example, an advertiser should not always be afraid to use negative contexts. These often have the highest rates of engagement, so there might be opportunity here, if the brand doesn’t choose to shy away from them.
Furthermore, there is research that indicates that brand perception is not negatively affected when an ad is shown in a negative context [2], like “Ukraine War” or “Hospital & Health” above. Rather, the results showed it was more important for the brand’s ad to be visible on websites from trusted publishers than avoiding content from those same publishers that might not be deemed as “safe”. The key of our technology is having insights and control over what kind of content and sites/apps to include or exclude from your campaign without sacrificing brand safety.
We’re ready to tailor each campaign for your specific brand safety needs while ensuring the most reach and engagement for your placed ad. This is just one of many reasons why you should future-proof your campaigns today, without dependency on cookies.
Research like Anton’s makes it possible for us to be at the forefront of contextual advertising. Don’t hesitate to reach out to us to set up your next campaign!
References
[1] CIPHE: A Framework for Document Cluster Interpretation and Precision from Human Exploration (Eklund et al., NLP4DH 2024) https://aclanthology.org/2024.nlp4dh-1.52/
[2] Häglund and Björklund. Should Advertisers Avoid Negative News? In Proceedings of the 2024 Ams Annual Conference, Coral Gables, Fl, Usa, May 22-24