SYSTEMS AND METHODS FOR PREDICTING CONTENT MEMORABILITY

Smart overview of the Invention

The patent application focuses on leveraging artificial intelligence (AI) and machine learning (ML) to assess the memorability of visual content. This involves predicting how likely a piece of visual content is to be remembered by viewers over time. The approach combines visual encoding and natural language processing (NLP) techniques to create a comprehensive analysis of content memorability.

Methodology

A key method involves transforming visual content into language tokens that represent it within a language space. These tokens are generated through two models: a visual encoding model and a verbalization model. The visual encoding model processes the visual data, while the verbalization model converts this data into a language format that can be further analyzed.

Role of NLP

To evaluate memorability, a natural language processing (NLP) model, such as a pre-trained large language model (LLM), is employed. This model is trained with memorability datasets, which enable it to interpret the language tokens derived from visual content and predict how memorable the content is likely to be.

Memorability Prediction

The system provides a memorability prediction by calculating the probability that the digital visual content will be remembered by an audience. This prediction encompasses both short-term and long-term recall, offering insights into how effectively the content can engage and remain in the minds of viewers.

Applications

Potential applications for this technology span various fields where understanding and enhancing content memorability is valuable. These include advertising, education, and media production, where knowing which elements of visual content are most impactful can guide design and presentation strategies.