Entity Extraction
How does Lexic know what's in your notes? Every piece of content you save is automatically analyzed to identify the people, organizations, concepts, and more that matter to you.
How It Works
When you save a note, Lexic uses GPT-3.5-turbo with function calling to intelligently identify and extract entities from your content. This isn't simple keyword matchingโit's context-aware AI that understands the meaning behind your words.
Confidence Scoring
Each extracted entity receives a confidence score. Only entities meeting the threshold (default: 0.7) are included, reducing noise and false positives.
Position Tracking
Lexic tracks the exact start and end character positions of each entity, enabling precise highlighting and navigation within your notes.
Context Awareness
The AI analyzes surrounding context (10-20 words plus the full sentence) to accurately classify entities and understand their role in your content.
Function Calling
Uses structured function calling for reliable, consistent extraction results with properly typed entity data.
Entity Types
Lexic recognizes eight distinct entity types, each designed to capture the key elements of professional knowledge work.
| Type | Examples |
|---|---|
| person | "Sarah Chen", "Dr. Williams" |
| organization | "Acme Corp", "Stanford University" |
| location | "San Francisco", "Building 4" |
| date | "Q3 2024", "next Tuesday" |
| money | "$50,000", "2.5M funding" |
| event | "board meeting", "product launch" |
| concept | "market fit", "technical debt" |
| technology | "React", "PostgreSQL" |
Personalization by Role
Entity extraction adapts to your professional role. Based on your profile, Lexic prioritizes the entity types most relevant to your work.
Product Manager
Prioritizes projects, timelines, and stakeholders
Engineer
Prioritizes systems, technologies, and dependencies
Researcher
Prioritizes concepts, citations, and methodologies
Executive
Prioritizes organizations, financials, and strategy
Designer
Prioritizes user needs, patterns, and feedback
Analyst
Prioritizes metrics, trends, and data sources
Consultant
Prioritizes clients, deliverables, and recommendations
Student
Prioritizes courses, assignments, and references
Tip: Set your role in your profile settings to get the most relevant entity extraction for your work.
Processing Details
Behind the scenes, Lexic handles the complexity of processing your content reliably and efficiently.
Automatic Text Chunking
Large inputs are automatically split into manageable chunks (approximately 3,000 characters each) to ensure thorough processing without overwhelming the AI model.
Retry Logic with Exponential Backoff
If a request fails, Lexic automatically retries with increasing delays between attempts, ensuring reliable extraction even during high load.
Batch Processing Support
Multiple notes can be processed together efficiently, making bulk imports and updates fast and cost-effective.
Deduplication
Extracted entities are automatically deduplicated, so "John Smith" mentioned five times in a note becomes a single entity reference with multiple occurrences tracked.
Cost Transparency
Entity extraction is included in note processing. When you save a note, you'll see the total word cost which includes extraction, embedding generation, and knowledge graph updates. No hidden fees or surprise charges.