Skip to main content

Connection Machine

Overview​

Connection Machine compares evidence sources in your Workspace and shows how they relate through shared identifiers and semantic matches. You can compare documents, folders, and images to find overlaps such as names, organizations, phone numbers, and locations. Outputs include a match table with traceable references and a differential view that highlights what is discovered thanks to their comparison. AI features improve coverage but require review, especially for semantic matches.

When to Use This Application​

  • You need to find overlaps across two document sets (for example two folders of reports).
  • You need to trace the same person or organization across multiple sources using shared identifiers.
  • You need to connect images or media by face or image similarity when text is incomplete.
  • You need to compare a focused source against a broader target (for example one file vs the entire workspace).
  • You need to identify what is new in an updated dataset using differential analysis.

Before You Begin​

  • Confirm the items you want to compare are available in a Workspace:
    • Documents
    • Folders
    • Images
  • Ensure the target set is defined (specific files, a folder, or the entire Workspace).
  • If you plan to use AI options, confirm AI features are enabled for your deployment.

Step-by-Step Walkthrough​

Step 1 — Select source and target​

Select or upload the source you want to compare. Then define the target scope. Typical options include:

  • Source: one or more files, a folder, or images
  • Target: specific files, a folder, or the entire Workspace

After selection, the app prepares inputs for identifier extraction and comparison.

Step 2 — Review extracted identifiers​

After you load source and target, the app extracts identifiers using standard parsing and shows them in tabs. Common identifier types include:

  • Names
  • Organizations
  • Dates
  • Email addresses
  • Phone numbers
  • Locations
  • Other structured entities

Review the identifier tabs before you run the comparison. Remove or correct items that are irrelevant to your investigation.

Step 3 — Enable AI deep analysis (optional)​

Enable AI Deep Analysis when documents are unstructured or when key identifiers are implicit. In this mode, the app analyzes full text to extract contextual identifiers that standard parsing may miss, such as:

  • Implicit names
  • Contextual organizations
  • Variations of the same entity
  • Less structured references

Use this option when you need more coverage before you run the comparison.

Step 4 — Enable Smart AI matching (optional)​

Enable Smart AI Matching when identifiers are inconsistent across sources. This option uses semantic matching to detect equivalences beyond exact identifier matches. Examples include:

  • Nicknames to formal names (for example “Johnny” → “John”)
  • Abbreviations (for example “NYC” → “New York City”)
  • Aliases and common misspellings Review these matches carefully. Treat them as suggested equivalences until you confirm them in the source content.

Step 5 — Run the comparison​

Start the comparison after you review identifiers and choose optional AI settings. The app compares the source and target using:

  • Extracted identifiers
  • AI Deep Analysis (if enabled)
  • Smart AI Matching (if enabled)

The app then produces a match table ranked by relevance.

Step 6 — Review match results​

Review the results table to see identifiers shared between source and target. By default, the app shows the most relevant matches first. Select View All to display all matches. Use filters to narrow results:

  • Filter by identifier type
  • Filter by label Open a match to view details. The details view shows:
  • Where the identifier appears in the source
  • Where it appears in the target
  • Context in each document

Use this view to validate each connection.

Step 7 — Run differential analysis (optional)​

Use Differential View to find identifiers that appear in the target but not in the source. This view is available in the details section of a matched identifier. Use differential analysis when you need to:

  • Compare document versions
  • Review newly acquired data
  • Identify newly surfaced connections in a broader dataset

Understanding the Output​

Connection Machine produces three main outputs:

  • Match table
  • A table of identifiers shared between the source and target. Each entry includes identifier type, identifier label, and references to where it appears.
  • Match details
  • A traceable view that shows the exact locations and surrounding context for each identifier in both source and target. Use this to confirm that a match is meaningful.
  • Differential view
  • A list of identifiers present in the target but not present in the source. Use this to detect new information when comparing updates.

If you enable semantic matching, treat suggested equivalences as leads. Confirm them by opening the referenced documents or media.

Tips for Best Results​

  • Define source and target carefully before you run the comparison. A broad target increases noise.
  • Review extracted identifiers before matching. Remove irrelevant identifiers to improve results.
  • Enable AI Deep Analysis for messy, unstructured, or narrative documents.
  • Enable Smart AI Matching when you expect aliases, abbreviations, or inconsistent naming.
  • Use filters to focus on a specific identifier type (for example phone numbers first).
  • Confirm semantic matches by opening the referenced sources before recording conclusions.

Known Limitations​

  • Standard parsing can miss contextual identifiers in unstructured text. Use AI Deep Analysis when coverage matters.
  • Smart AI Matching suggests semantic equivalences, not confirmed facts. You must validate matches in the source material.
  • Match quality depends on how well identifiers are extracted from the selected sources.