Skip to content

Schema-Agnostic Database Access

A query-mediation capability that maps user terminology and structure onto an unfamiliar or changing database schema, constructs executable interpretations, and exposes ambiguity without requiring native vocabulary knowledge.

Version
v1 · 2026-09-28 · History
Domain-specific #
11897
Domain group
Applied Sciences & Engineering
Origin domain
Computer Science & Software Engineering
Subdomains
Semantic Data Access, Databases, Knowledge Graphs → Computer Science & Software Engineering
Aliases
Schema-agnostic querying, Schema-agnostic database querying, Schema-independent querying

Core Idea

Schema-agnostic access addresses the semantic gap between an information need and a database's representational choices. Users describe desired entities and relations in their own vocabulary; the system discovers candidate mappings to tables, properties, predicates, joins, or graph paths.

The result must be more than keyword search. Mapped concepts are composed into an executable structured query, competing interpretations are ranked or clarified, and answers retain provenance to the selected schema elements. Natural language, keywords, and relaxed structured syntax are possible front ends to this mediation task.

Structural Signature

Sig role-phrases:

  • User information need — Expresses entities, relations, and constraints in user language. It is required input. Counterfactual: A fully native structured query already assumes schema knowledge.
  • Unknown data representation — Supplies tables, predicates, properties, or graph paths not presumed known to the user. It is defining gap. Counterfactual: If the user and database share a fixed vocabulary, ordinary query tools suffice.
  • Semantic matching — Proposes correspondences between user expressions and data elements. It is mediation core. Counterfactual: String matching alone cannot resolve many vocabulary and structural differences.
  • Structural interpretation — Assembles matched elements into joins, paths, filters, and answer shape. It is required composition. Counterfactual: A bag of matching records may not satisfy a relational information need.
  • Candidate ranking and explanation — Handles ambiguity and makes mapping choices inspectable. It is uncertainty control. Counterfactual: One silent mapping can return plausible but semantically wrong answers.
  • Execution and validation — Runs interpretations against the source and checks answer coherence. It is output gate. Counterfactual: A linguistic paraphrase without executable grounding is not database access.

What It Is Not

  • It is not the same as a schema-less database.
  • It is not merely hiding SQL behind a form whose fields mirror the schema.
  • It is not unrestricted document keyword search.
  • It does not eliminate schema; it interprets and mediates it.
  • Closest near-miss. A natural-language interface describes the input mode; it is schema-agnostic only when it also resolves unfamiliar dataset vocabulary and structure.

Scope of Application

  • Knowledge graphs. Maps user concepts onto heterogeneous predicates and paths.
  • Data integration. Queries sources with independent vocabularies and structures.
  • Natural-language interfaces. Grounds utterances in executable database semantics.
  • Exploratory analytics. Lowers the entry cost where schemas are large or rapidly changing.

Clarity

State which schema knowledge is not required, which metadata or ontologies the system uses, how relations are composed, and how ambiguity is surfaced. Vocabulary independence is a degree, not an all-or-nothing property.

Manages Complexity

The abstraction separates user intent from data representation while concentrating uncertainty in a mediation layer. It makes heterogeneous data usable but demands explicit ranking, provenance, and correction mechanisms.

Abstract Reasoning

  1. Parse entities, relations, constraints, and answer type from the request.
  2. Retrieve candidate schema elements using lexical and semantic evidence.
  3. Compose candidates into structurally valid interpretations.
  4. Rank, clarify, or expose alternatives before execution.
  5. Execute and return answers with mapping provenance.

Knowledge Transfer

The pattern transfers across relational, graph, and federated data when a formal query can be grounded. It does not transfer directly to unstructured retrieval without a representational mapping problem.

Examples

Canonical

A user asks for rivers crossing capitals without knowing graph predicates; the system maps river, crossing, capital, and country relations to candidate paths, ranks interpretations, executes the selected graph query, and shows the mapping.

Mapped back: need → user vocabulary; match → ontology predicates; structure → multi-hop path; control → ranked explainable mapping.

Applied / In Practice

A search box returns rows containing the word capital but cannot connect cities, countries, and rivers; it is keyword retrieval rather than schema-agnostic structured access.

Mapped back: lexical match → present; relation composition → absent; verdict → near-miss.

Structural Tensions

T1 — User Abstraction versus Semantic Control. Hiding schema lowers access cost but can conceal consequential mapping errors.

Diagnostic: Can users inspect, constrain, or correct the interpretation?

T2 — Broad Coverage versus Precise Interpretation. Flexible matching covers heterogeneous sources while multiplying plausible schema paths.

Diagnostic: How are ambiguity, confidence, and false correspondences managed?

Structural–Framed Character

Schema-Agnostic Database Access is mixed: mapping and execution are structural, while lexical meaning and relevance are user- and dataset-framed.

Structural Core vs. Domain Accent

The skeleton is semantic mediation from intent to executable structure. Databases supply schemas, joins, graph paths, query languages, metadata, and provenance.

This entry is a kind of Semantic translation.

  • Approved root. No current node entails this full vocabulary-to-query mediation loop.

  • Related — ontology alignment, semantic parsing, and data integration. They supply correspondence, interpretation, and source combination.

Relationships to Other Abstractions

Local relationship map for Schema-Agnostic Database AccessParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Schema-AgnosticDatabase AccessDOMAINDomain-specific abstraction: Semantic translation — is a kind ofSemantictranslationDOMAIN

Current abstraction Schema-Agnostic Database Access Domain-specific

Parents (1) — more general patterns this builds on

  • Schema-Agnostic Database Access is a kind of Semantic translation Domain-specific

    Schema-Agnostic Database Mediation is Semantic Translation from user concepts and structures into executable terms of an unfamiliar or changing database schema.

Hierarchy paths (2) — routes to 2 parentless roots

Neighborhood in Abstraction Space

Schema-Agnostic Database Access sits in a moderately populated region (41st percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Matrices, Measures & Numeric Structures (30 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • Schema-less database. Tell: Allows flexible stored records but still has fields and conventions users may need to know.
  • Natural-language query. Tell: Specifies an input language but may rely on one known schema.
  • Keyword search. Tell: Retrieves lexical matches without necessarily composing relations.
  • Object–relational mapping. Tell: Maps program objects to a known relational schema.

References

  • Frozen Wikipedia discovery revision: https://en.wikipedia.org/wiki/Schema-agnostic_databases (revision 1313544210).
  • Preserved source candidate: http://andrefreitas.org/papers/thesis_schema_agnostic.pdf
  • Preserved source candidate: http://andrefreitas.org/papers/schema_agnostic_i_final_version.pdf
  • Preserved source candidate: http://dbpedia.org/resource/
  • Preserved source candidate: http://dbpedia.org/property/
  • Preserved source candidate: http://dbpedia.org/ontology/
  • Preserved source candidate: http://www.w3.org/2004/02/skos/core#
  • Preserved source candidate: http://www.w3.org/1999/02/22-rdf-syntax-ns#
  • Preserved source candidate: http://andrefreitas.org/papers/preprint_iwcs_2015.pdf

The frozen Wikipedia revision is discovery provenance. The retained source set was reviewed for identity, formal or operational relation, and scope. The encyclopedia's structural synthesis is bounded to those claims; a thin authority surface is recorded as a nonblocking source-strengthening repair rather than concealed.