Novel Friend Recommendations Based on User-Generated Contents in Online Social Media Sites

Abstract

The number of users joining and contributing to Social Networking Sites (SNSs) has steadily increased. Users share personal experiences such as food, shopping, travel, and leisure activities, among others, and also leave comments. The significant volume of User-generated content (UGC) offers ample data for constructing a model that accurately reflects users’ sentiments towards various locations, interests, and personalities. The model should also be adaptable and scalable. These attributes can affect friend matches via the computer-supported social matching process. Regrettably, prior research in this domain omitted the use of extracted meta-text features in developing friend recommendation systems. This study proposes a text analytic framework the author applies to UGCs on SNSs. Extracting interests and personality features from UGCs enables text-based friend recommendations. The experiment results demonstrate that textbased features can enhance recommendation performance.