Sentiment and Emotion Detection

Tisane Labs reads a passage and returns the sentiment and emotion expressed inside it. When a new review or ticket arrives, a Beam agent sends the text for analysis, reads the polarity result, and applies your rule for what counts as negative. Records that cross the threshold get tagged and the owning team is notified, while positive and neutral ones are filed quietly. Messages with mixed or borderline scoring, or text the model returns low confidence on, go to a person so the tone is judged by a human before any reply is drafted.

Abuse and Content Moderation

The moderation function inspects text for hate speech, profanity, and other policy breaches, with an explanation of why each was raised. A Beam agent runs incoming comments through it, reads the categories returned, and applies your community rule about what is allowed. Clear violations are hidden or queued and the moderator channel is notified, while clean posts pass through. Anything near the policy line, sarcasm, or content in a dialect the model is unsure about is routed to a human reviewer, who makes the final call and keeps the account record accurate.

Entity and Language Extraction

Language detection and entity extraction identify the language of a passage and pull out the names, places, and organizations mentioned in it. On a new document, a Beam agent detects the language first, then reads the entities and writes them onto the record as structured fields for search and routing. Your rule decides which entity types matter and where each one is stored. Passages mixing several languages, or where an entity is ambiguous between a person and a place, are handed to someone to label correctly before the record is treated as complete.