Skip to main content

Entity Extraction

Automatically extract and analyze people, organizations, money amounts, and other key entities from your case documents.

What is Entity Extraction?​

Entity extraction uses AI to automatically identify and categorize important information in your documents:

Entity TypeExamples
PeopleJohn Smith, Dr. Sarah Martinez, Michael Chen
OrganizationsAcme Corp, Global Holdings Inc, Department of Justice
Money$500,000, $1.5 million, 125,000 shares
DatesAugust 7, 2024, March 2024
Emailsjohn@company.com, contact@acme.com
Phones(555) 123-4567

How It Works​

BillionLens uses Grounded Extraction with citation verification:

  1. Document Search - AI searches all indexed documents
  2. Entity Detection - Identifies entity mentions with citations
  3. Verification - Each entity is verified against source documents
  4. Description Generation - AI creates contextual descriptions
  5. Alias Detection - Groups name variations (e.g., "Mike" = "Michael Chen")
  6. Deduplication - Merges duplicate entities intelligently
Citation-Required

Every extracted entity must be verified in source documents. This prevents hallucination and ensures accuracy for e-discovery.

Running Entity Extraction​

Extract Entities​

  1. Navigate to your case
  2. Click Entities card
  3. Click Extract Entities button
  4. Wait for extraction to complete
  5. View extracted entities organized by type

Progress Tracking​

During extraction, you'll see:

  • Progress percentage
  • Current extraction phase
  • Entity counts as they're discovered

Viewing Entities​

Entity Tabs​

Entities are organized by type:

TabDescription
AllAll entity types combined
PeoplePerson names
OrganizationsCompanies, agencies, institutions
MoneyDollar amounts and financial values
DatesImportant dates
EmailsEmail addresses

Entity Details​

Click the Context link on any entity to see:

  • Description - AI-generated explanation of the entity's role
  • Also Known As - Alias names (for people)
  • Source Documents - Which files mention this entity

Example Entity​

John Smith (Person)

  • Description: "John Smith is the Chief Technology Officer at Acme Corp, responsible for overseeing product development and engineering teams."
  • Also Known As: J. Smith, Johnny

Alias Detection & Deduplication​

Automatic Alias Detection​

BillionLens automatically detects when different names refer to the same person:

Examples:

  • "John Smith" = "Johnny Smith" = "J. Smith"
  • "Michael Chen" = "Mike Chen" = "M. Chen"
  • "Dr. Martinez" = "Sarah Martinez" = "S. Martinez"

How It Works​

The AI uses several strategies:

  • Middle names - "John Michael Smith" matches "Michael Smith"
  • Nicknames - "Mike" is recognized as nickname for "Michael"
  • Abbreviations - "J. Smith" matches "John Smith"
  • Typos - "Jhon" matches "John"

Merged Entities​

When aliases are detected:

  • One canonical name is chosen (usually the most complete)
  • Other variations are stored as "Also Known As"
  • All document references are preserved

Entity Descriptions​

AI-Generated Descriptions​

Each entity gets a contextual description based on document content:

Good description:

"Michael Chen holds the roles of COO/Representative and Chief Executive Officer, and is also listed as a founder and initial member of the board of directors."

Not just quotes:

Descriptions explain the entity's role and significance, not just raw quotes from documents.

Description Sources​

Descriptions are generated by:

  1. Finding all mentions of the entity in documents
  2. Understanding the context of each mention
  3. Synthesizing a concise explanation
  4. Verifying against source citations

Searching Entities​

Filter by Type​

Click the entity type tabs to filter:

  • View only People
  • View only Organizations
  • View only Money amounts
  • etc.

Use the search box to find specific entities:

  • Search for "Smith" to find all Smith entities
  • Search for company names
  • Search for specific amounts

Use Cases​

Building Witness Lists​

Extract PERSON entities to identify:

  • Potential witnesses
  • Parties involved
  • Key decision makers
  • People who signed documents

Identifying Key Players​

Find ORGANIZATION entities to see:

  • Companies involved
  • Law firms
  • Government agencies
  • Banks and financial institutions

Financial Analysis​

Extract MONEY entities to:

  • Track investment amounts
  • Identify financial transactions
  • Find payment records
  • Understand ownership stakes

Contact Discovery​

Find EMAIL entities to:

  • Build contact lists
  • Identify key communicators
  • Track correspondence patterns

Accuracy & Verification​

Citation-Required Extraction​

Unlike simple NER (Named Entity Recognition), BillionLens uses grounded extraction:

Traditional NERBillionLens Grounded Extraction
Pattern matchingAI with document search
May hallucinateCitation-required
No contextFull document context
No descriptionsAI-generated descriptions

Verification Process​

Every entity must pass verification:

  1. Entity name must appear in source documents
  2. Citation must be retrievable
  3. Description must be grounded in document text

Rejected Entities​

Entities that can't be verified are rejected:

  • Prevents hallucination
  • Ensures e-discovery accuracy
  • Logs rejected entities for review

Best Practices​

Review Important Entities​

  • Manually verify critical entity extractions
  • Cross-reference with source documents
  • Check for missing entities

Use Descriptions​

  • Read AI-generated descriptions for context
  • Descriptions explain relationships and roles
  • More useful than just names

Combine with Chat​

  • Use entity extraction to find candidates
  • Use AI chat to explore relationships
  • Ask follow-up questions about specific entities

Re-extract When Needed​

  • Re-run extraction if you add new documents
  • Previous entities are cleared before re-extraction
  • Ensures consistency

Next: Learn about Timeline Visualization.