Research in the field of educational videos and the contribution of data mining to education can affect the instructors’ approach to learning. This particular study focuses on online educational videos and more specifically on their speakers. Initially a survey is conducted related to the popularity of educational videos on the YouTube which are then divided into two categories the more popular and the less popular. Then the characteristics related to language are extracted from the transcript of the speakers and after a clustering procedure the differences between the two categories are stated. The characteristics related to the language of the speakers of the popular videos present very interesting results. That is, the pace of speaking is faster and the complexity off the sentences is higher than the ones in the less popular videos.