Named entity recognition

Lettria reads raw text and labels the people, places, organizations, and terms inside it. A Beam agent feeds incoming documents or messages into the model, reads the returned entities, and applies your rule to tag each record and file it under the right project. When the model returns low confidence or conflicting labels, the agent sets the record aside and flags it for a reviewer. A person confirms the labels on ambiguous text and decides how new entity types should be handled as fresh material arrives.

Sentiment scoring

Lettria scores text for sentiment and emotional tone across a whole document set. A Beam agent watches for new feedback or survey responses, reads the sentiment output, and applies your threshold rule to route negative cases to the right team and mark clearly positive ones as resolved. Responses that score near the neutral boundary, or that mix praise with complaint, are handed to a person for a closer read. The agent updates each record with its score so the team can watch the trend at a glance. Humans decide when a pattern of negative scores warrants escalation to a manager for a wider look.

Text classification

Lettria sorts text into the categories your team defines, such as topic, intent, or department. A Beam agent reads each incoming item, requests a classification, and applies your rule to move the record into the matching queue or workflow. When an item falls between categories or matches none with confidence, the agent notifies a person rather than guessing at a label. A reviewer confirms the category on edge cases and refines the label set as new kinds of text arrive over time. The agent then follows the updated definitions and keeps sorted records in step with the current rules.