Overview

Extracts entities from conversations to build and update user profiles, track preferences, and capture facts. It supports disambiguation and relationship mapping, plus configurable schemas and extraction rules to tailor personalization.

How This Skill Works

The skill identifies 5 extraction types: Named Entities, User Preferences, Facts, Temporal data, and Relationships. It offers configurable options such as extraction model selection, entity schema definitions, confidence thresholds, update policies, and storage backend. It leverages LangChain and SpaCy (optional) and can plug in custom extraction models to operate across long-term memory and conversational persona design.

When to Use It

• During multi-turn chats to persist and update a persistent user profile over time
• When personalizing recommendations or content based on expressed preferences and facts
• To disambiguate entities that recur across sessions and resolve conflicting data
• When building entity relationship graphs linking people, places, and events
• During configuration and governance to set schemas, thresholds, and privacy policies

Quick Start

1. Step 1: Define entity schemas (named entities, preferences, facts, temporal, relationships) and set confidence thresholds
2. Step 2: Enable extraction in the conversation layer and connect to a storage backend for profile updates
3. Step 3: Run regular profile validation, handle conflicts, and enforce privacy policies

Best Practices

• Define clear entity schemas before extraction to ensure consistency
• Implement conflict resolution and entity disambiguation strategies
• Apply confidence scoring and review low-confidence extractions
• Regularly validate and refresh profiles to reflect new data
• Incorporate privacy-by-design: enforce data minimization and access controls

Example Use Cases

• Retail assistant builds a shopper profile by extracting product preferences, sizes, and brands from chat history
• Travel planner records dates, destinations, and scheduling preferences to suggest itineraries
• Customer support bot links a user to organizations and roles mentioned in tickets for faster routing
• News app tailors feeds by tracking topics of interest and followed authors from conversations
• Enterprise bot maps relationships between colleagues, departments, and partner companies

Frequently Asked Questions