A Computational Model to Predict the Memorability of Web-pages

Abstract

In today's digital world, websites are the main point of interaction for a wide range of online activities. As a result, website memorability has become an important topic of discussion. In order to stand out in a highly competitive environment where users are constantly bombarded with information, a website's ability to be memorable is crucial to its success. This study focuses on the development of an automatic computational model for predicting the memorability of a web page. To achieve this, the objects within a web page were identified and their memorability scores were calculated using the ResNet-18 convolutional neural network. The final memorability score of the web page was computed by taking a weighted sum of the areas occupied by these objects on the web page, along with their memorability scores. For the empirical study, 30 web pages from different applications were used to train and test our proposed model. Our model can predict web page memorability with a mean absolute error of 0.077 on a normalized scale of 1.