The AlphaSense Sentiment Score is calculated as the net sum of positive and negative discourse in public company earnings call transcripts. It is powered by the AlphaSense Sentiment language model that applies deep domain-aware semantic reasoning to financial disclosure commentary and dialogue. It identifies significant shifts in executive tone and intent down to the sentence level, highlighted and auditable in all transcripts within the AlphaSense platform. This model delivers a normalized sentiment score (-100 to 100) that allows the tracking and identification of sentiment shifts based on what executives across thousands of public companies are discussing on earnings calls.