Skip to main content

Under the Hood

1. Building the Knowledge Graph​

Core Components​

Rezolve’s knowledge graph integrates three pillars of enterprise data:

  • Content: Documents, messages, tickets, and other assets from apps like Slack, Salesforce, or Google Drive.

  • People: Employee identities, roles, teams, and departments.

  • Activity: User interactions (clicks, edits, comments) and content lifecycle (creation/modification dates).

Construction Process​

Data Ingestion via Connectors​

Rezolve uses 100+ pre-built connectors to pull structured and unstructured data from enterprise apps (e.g., Slack, Jira, Confluence).

Connectors normalize data into a unified format and respect source permissions (e.g., private Slack channels are excluded from unauthorized users’ search results).

Entity Extraction & Relationship Mapping​

Named Entity Recognition (NER): Identifies entities (people, projects, acronyms) and maps their relationships (e.g., "John authored Q4 Sales Report").

Semantic Understanding: Uses ML models (likely BERT or similar) to infer implicit relationships (e.g., "FY24" ≈ "2024 fiscal year"). TODO

Activity Signal Integration​

Tracks user behavior (clicks, searches, edits) to infer document popularity, team relevance, and context.

Adaptation to Organizational Language​

Rezolve’s Scholastic system fine-tunes language models on each customer’s corpus to understand internal jargon and communication patterns.

Query Processing​

Utilizes vector embeddings and semantic search to find the most relevant results.

Knowledge Graph Traversal​

Follows relationships between entities to provide contextually rich results.

Result Ranking​

Relevance​

Ranks results based on semantic similarity and user activity signals.

Contextual Relevance​

Prioritizes results based on team relevance and document popularity.

Result Presentation​

Knowledge Cards​

Displays concise summaries of documents, messages, and tickets with relevant metadata.

Contextual Clusters​

Groups results by topic and team to provide a clear overview of the search results.

3. Knowledge Graph Management​

Content Lifecycle​

Tracks document and message lifecycle events (creation, modification, deletion).

Entity Management​

Maintains up-to-date entity information and relationships.

Activity Management​

Tracks user interactions and activity signals.

Corpus Adaptation​

Continuously updates language models to adapt to organizational language and communication patterns.

4. Knowledge Graph Maintenance​

Content Lifecycle​

Tracks document and message lifecycle events (creation, modification, deletion).

Entity Management​

Maintains up-to-date entity information and relationships.

Activity Management​

Tracks user interactions and activity signals.

Corpus Adaptation​

Continuously updates language models to adapt to organizational language and communication patterns.

Knowledge Refinement​

Graph Completion​

Predicts missing links (e.g., auto-linking a new employee to their team based on email domain).

Noise Reduction​

Filters outdated or low-engagement content (e.g., deprecated Confluence pages).

Permission Sync:

Mirrors access controls from source systems (e.g., revoked Slack access removes related content from search)