Lexicon-based sentiment scoring for text, returning label, scores, and positive/negative word hits.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.