Ranking documents based on user behavior and/or feature data
Topics: AI (Deep Learning), Backlinks, E-E-A-T, Jeff Dean, Ranking, User Signals
The patent describes a system for ranking documents based on user behavior and feature data of links. It introduces a “reasonable surfer model” that assigns different probabilities to links being followed based on their characteristics. The system generates a model using data about how users interact with links (like click patterns) and features of those links (such as font size, position, etc.). This model is then used to assign weights to links and ultimately rank documents, which helps improve search result quality.