AI knowledge graph generator from document evidence
Turn documents into a structured map of the people, companies, organizations, assets, concepts and other entities in the record. Continua resolves supported aliases, extracts evidence-backed relationships, preserves uncertainty and cites the source behind each material node and connection.
No credit card required · Uploaded files deleted after the task
What is an AI knowledge graph generator?
An AI knowledge graph generator identifies entities in unstructured documents and maps the relationships between them. Continua creates canonical entity records, resolves name variations when the evidence supports a match, and includes only relationships explicitly stated or unambiguously established by the uploaded files. Ambiguous identities and links remain unresolved rather than being invented.
See the relationship and the evidence behind it
Illustrative entity and relationship records
| Source entity | Relationship | Target entity | Valid from | Evidence | Status |
|---|---|---|---|---|---|
| Northstar Holdings LLC | owns 80% of | Delta Operations Inc. | January 2024 | Ownership schedule, p. 6 | Supported |
| N. Holdings | alias of | Northstar Holdings LLC | Not applicable | Defined terms, p. 2 | Supported |
| Delta Operations Inc. | guarantees obligations of | Harbor Services LLC | March 2024 | Guaranty, p. 1 | Supported |
| “Northstar” | may refer to | Northstar Holdings LLC or Northstar Fund II | Not established | Email thread, p. 4 | Ambiguous—do not merge |
Graph finding: “N. Holdings” is defined as Northstar Holdings LLC and can be merged into the canonical entity. The standalone name “Northstar” is used inconsistently and should remain unresolved until additional evidence identifies the intended entity.
This is an illustrative example. A Continua graph keeps entity and relationship findings linked to their source documents and locations.
Names are easy to extract. Identity and relationships are harder.
The same organization may appear under a legal name, trading name, abbreviation and typo. Two people may share a surname. A subsidiary may be described as an affiliate in one document and a guarantor in another. A relationship may begin on an effective date that differs from the signature date.
The useful question is not only “Which names appear in these files?” It is:
Which entities are actually the same, how are they connected, when did the relationship apply and what evidence supports the link?
Continua separates entity identification, alias resolution and relationship extraction so uncertain matches do not become false connections.
What the document knowledge graph includes
Canonical entities
Create one record for each supported person, organization, asset, location, agreement, product, concept or other entity relevant to the investigation.
Entity types and descriptions
Classify entities and add concise descriptions grounded in the record, helping reviewers distinguish similarly named people or organizations.
Alias resolution
Merge legal names, abbreviations and known variations only when the evidence supports that they refer to the same entity.
Evidence-supported relationships
Extract the source entity, relationship and target entity for ownership, employment, control, contracting, guarantees, supply, participation and other documented connections.
Relationship dates
Capture valid-from or relevant date ranges when the evidence states or supports when a relationship began, changed or ended.
Uncertainty and ambiguity
Keep possible identity matches or relationships separate when the record does not provide enough evidence to resolve them.
Source citations
Link material entities, aliases and relationships to the exact documents and locations that establish them.
Ask the complete evidence set one graph question
Build a knowledge graph from all uploaded files. Identify the material people, companies, organizations, assets, agreements, locations and concepts. Create a canonical entity record with type and evidence-based description. Resolve aliases only when the documents support the match. Extract each source entity, relationship, target entity and valid-from or date range when available. Cite every material entity and relationship. Preserve ambiguous identities, conflicting relationships and uncertainty instead of inventing or over-merging links.
Knowledge graph use cases
Corporate ownership and control
Map parents, subsidiaries, owners, directors, guarantors and related entities across corporate records, contracts and disclosures.
Contract parties and obligations
Connect agreements, counterparties, affiliates, assets, assignments, guarantees and the entities bound by specific provisions.
Due diligence
Identify relationships hidden across a data room and flag inconsistencies in entity names, ownership statements or related-party descriptions. See AI due diligence.
Investigations and claims
Map people, events, communications, organizations and assets across reports, correspondence and case files.
Research landscapes
Connect authors, institutions, methods, concepts, populations and findings across a literature collection.
Projects and systems
Map components, teams, vendors, dependencies, incidents and decisions when the relationships are spread across technical and operational records.
From documents to a cited relationship map
- 01Upload the relevant record set. Include sources that identify entities, aliases, roles, dates and relationships.
- 02Define the graph scope. Specify the entity types and relationships that matter to the investigation.
- 03Extract and resolve entities. Continua creates canonical records and tests supported name variations.
- 04Map relationships and uncertainty. Review connections, dates, conflicting sources and ambiguous identities.
- 05Open evidence and export the report. Verify material nodes and edges at their sources before using the graph in professional analysis.
Knowledge graph vs. a list of extracted names
| Name or entity list | Continua knowledge graph |
|---|---|
| Captures mentions | Creates canonical entity records |
| May duplicate aliases | Resolves supported name variations |
| Shows entities independently | Maps evidence-backed relationships |
| Often ignores time | Captures relationship dates when available |
| Can over-merge similar names | Preserves ambiguous identity matches |
| Provides limited provenance | Cites material entities and connections |
Questions a document knowledge graph can answer
- Which legal names and aliases refer to the same entity?
- Who owns, controls, employs or guarantees whom?
- Which parties are connected to an agreement, asset or event?
- When did a relationship begin, change or end?
- Which relationship is stated directly, and which remains uncertain?
- Are entity descriptions inconsistent across documents?
- Which people or organizations connect otherwise separate evidence?
- What source establishes each material connection?
Supported files, security and evidence discipline
Continua can analyze PDF, DOCX, XLSX and CSV files together. Uploaded files are deleted after the task and are not used to train AI models. Material graph findings remain linked to supporting source evidence.
The graph reflects the uploaded record, not external truth. An absent relationship means it was not established in the provided files; it does not prove no relationship exists. Ambiguous identities should remain separate until evidence supports a merge.
Related Continua workflows
Frequently asked questions
Can Continua create a knowledge graph from multiple documents?
Yes. It can identify entities and relationships across the uploaded file set, allowing a connection stated in one document to be evaluated alongside aliases or context found elsewhere.
Can the AI merge aliases and name variations?
Yes, when the evidence supports that the names refer to the same entity. Ambiguous or conflicting identity matches should remain separate.
Will Continua invent relationships?
No. The workflow is designed to include relationships explicitly stated or unambiguously established by the uploaded evidence and to preserve uncertainty when support is insufficient.
Can relationships include dates?
Yes. A valid-from date or date range can be captured when the source states or supports when the relationship began, changed or ended.
What types of entities can be mapped?
Depending on the record, entities can include people, companies, organizations, assets, agreements, locations, products, systems, events and concepts.
Does an absent link mean no relationship exists?
No. It means the relationship was not established in the uploaded file set. Additional documents or external research may be needed.
See the network hidden inside your documents
Upload the complete evidence set and get a cited map of entities, aliases, relationships, dates and unresolved identity questions.
No credit card required · Zero document retention
Start freeRun Your First Investigation Free
Attach the file pile, ask the hard question, get back a cited report with every conflict flagged. No credit card required.