What it does
"Activation Potential: Text” measures the arousal potential of text within an image or video, using a deep language model. It evaluates how stimulating and activating the text is likely to be for viewers, based on psychological principles of arousal and attention.
Why it matters
Arousal in marketing, especially through text, is critical for capturing attention, increasing emotional engagement, and facilitating deeper message processing. It influences the viewer's alertness and readiness to respond, making assets more memorable and impactful. Assets with text that elicits strong emotional responses are more likely to capture attention and encourage action, such as purchase or sharing. However, it is important to balance the level of excitement to avoid overwhelming the viewer.
How it works
Text is extracted from images or videos using a text recognition model, and words are filtered based on confidence level and character length to ensure clarity and semantic significance. The model then scores each by determining its potential to stimulate and engage viewers. Words are weighted by their relative size, as larger words have a greater impact on overall arousal. The final score combines these individual word scores into an average that reflects the text’s overall arousal potential.
How to improve
- Use emotionally charged words that are known to increase arousal and capture attention.
- Ensure that key messages are prominently displayed and of significant size to increase their impact.
- Balance the text’s arousal potential with the asset’s overall design to maintain viewer engagement without causing discomfort.
AI models used
The model was initially trained to understand text in general, using a large corpus of text from multiple sources, comprising over 100 million words. It was then fine-tuned to predict the arousal potential of human-rated words (>20,000 words). The resulting model has an accuracy of 91%. Text recognition is based on a world-class OCR model with 98% accuracy, capable of recognizing a wide range of fonts and styles in diverse visual environments.