You are working with customer feedback from digital service interactions, call center notes, and survey comments. The goal is to turn free text into sentiment labels so analysts can track where customers are frustrated, satisfied, or neutral across the experience. The text is messy, often short, and may include insurance terms, abbreviations, claim references, and mixed signals in the same comment.
How would you classify customer comments by sentiment?
Short survey comments and longer service notesMixed polarity in one commentInsurance vocabulary such as claim, deductible, premium, endorsementPossible identifiers that should be masked before training