This study is based on the specific corpus of tutoring session transcripts, and whether its conclusions can generalize to other text domains such as news, legal, and medical texts requires further verification.