Lecture videos are amongst the most widely used instructional methods within present Massive Open Online Courses (MOOCs) and other digital educational platforms. As the main form of instruction, student engagement behaviour, including interaction with videos, directly impacts the student success or failure and accordingly, in-video dropouts positively correlate with dropout from MOOCs. The primary focus of previous video learning analytics studies is on analyzing video interaction behaviour using "explicit" factors (i.e., views). Limited research studies focus on "implicit" video learning analytics (e.g., pause, seek, content type) and their impact on student success, with existing studies addressing video interactions and their relationship with visual transitions. We aim to explore the association between video interactions and non-visual (i.e., verbal) content. This research focuses on language and discourse features of lecture video contents. We conduct a fine-grained analysis of 3.4 million video interactions across two AdelaideX MOOCs -- Programming (Code101x) and Cyberwar, Surveillance and Security (Cyber101x). According to our results, a number of discourse features (e.g., lexical diversity and causal connectives) demonstrate a statistically significant correlation with video interactions. We present insights regarding educational video design implications based on discourse processing theories.