With the proliferation of Web 2.0 technologies, folksonomy which is also known as social tagging or collaborative tagging is widely used by learners to annotate and categorize their learning resources. In a folksonomy system, the tags are added by learners to the learning resources, hence the tags are often ambiguous, overly personalised and imprecise. In addition, conjugated words, compound words and nonsense words may be used in tagging and shared among a group of learners. This has resulted in an uncontrolled and chaotic set of tagging terms that cause learning resources searching, reuse and sharing to become ineffective. In this paper, we present a content-based approach which automatically generates tags from a learning resource using Part-Of-Speech Tagging and K-Means Clustering techniques. The generated tags are more precise and unambiguous which can improve learning resources searching, reuse and sharing among learners.