Context Inference Using Correlation in Human Behaviour

Abstract

Abstract-Context-aware information retrieval received a significant amount of attention last years. This paper addresses some of the challenges in context acquisition. It is focused on a method for inference of unavailable contextual information using machine learning. The method for context inference is based on observed behaviour of individual user and virtual communities of similar users. We work with with contextual information such as location, time or weather in the domain of news recommending. We discuss the role of user behaviour and its significant impact on the actual information need, which directly influences information retrieval or recommendation process. In experiments we demonstrate the impact of inferred context on the recommendation process and its precision and recall. We use behaviour of news readers to predict their interest in news. We present context-aware recommendation which is supported by our method for context inference.

Context-Based Satisfaction Modelling for Personalized Recommendations

“…In the real world the recommendation always depends heavily on user's context [14]. Approaches that consider the user's context represent an enhancement of above described approaches.…”