Hubbry Logo
Semantic data modelSemantic data modelMain
Open search
Semantic data model
Community hub
Semantic data model
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Contribute something
Semantic data model
Semantic data model
Semantic data model
View on Wikipedia
from Wikipedia
The relationship of "Semantic data models" with "physical data stores" and "real world".[1]

Key Information

A semantic data model (SDM) is a high-level semantics-based database description and structuring formalism (database model) for databases. This database model is designed to capture more of the meaning of an application environment than is possible with contemporary database models. An SDM specification describes a database in terms of the kinds of entities that exist in the application environment, the classifications and groupings of those entities, and the structural interconnections among them. SDM provides a collection of high-level modeling primitives to capture the semantics of an application environment. By accommodating derived information in a database structural specification, SDM allows the same information to be viewed in several ways; this makes it possible to directly accommodate the variety of needs and processing requirements typically present in database applications. The design of the present SDM is based on our experience in using a preliminary version of it. SDM is designed to enhance the effectiveness and usability of database systems. An SDM database description can serve as a formal specification and documentation tool for a database; it can provide a basis for supporting a variety of powerful user interface facilities, it can serve as a conceptual database model in the database design process; and, it can be used as the database model for a new kind of database management system.

In software engineering

[edit]

A semantic data model in software engineering has various meanings:

  1. It is a conceptual data model in which semantic information is included. This means that the model describes the meaning of its instances. Such a semantic data model is an abstraction that defines how the stored symbols (the instance data) relate to the real world.[1]
  2. It is a conceptual data model that includes the capability to express and exchange information which enables parties to interpret meaning (semantics) from the instances, without the need to know the meta-model. Such semantic models are fact-oriented (as opposed to object-oriented). Facts are typically expressed by binary relations between data elements, whereas higher order relations are expressed as collections of binary relations. Typically binary relations have the form of triples: Object-RelationType-Object. For example: the Eiffel Tower <is located in> Paris.

Typically the instance data of semantic data models explicitly include the kinds of relationships between the various data elements, such as <is located in>. To interpret the meaning of the facts from the instances, it is required that the meaning of the kinds of relations (relation types) be known. Therefore, semantic data models typically standardize such relation types. This means that the second kind of semantic data models enables that the instances express facts that include their own meanings. The second kind of semantic data models are usually meant to create semantic databases. The ability to include meaning in semantic databases facilitates building distributed databases that enable applications to interpret the meaning from the content. This implies that semantic databases can be integrated when they use the same (standard) relation types. This also implies that in general they have a wider applicability than relational or object-oriented databases.

Overview

[edit]

The logical data structure of a database management system (DBMS), whether hierarchical, network, or relational, cannot totally satisfy the requirements for a conceptual definition of data, because it is limited in scope and biased toward the implementation strategy employed by the DBMS. Therefore, the need to define data from a conceptual view has led to the development of semantic data modeling techniques. That is, techniques to define the meaning of data within the context of its interrelationships with other data, as illustrated in the figure. The real world, in terms of resources, ideas, events, etc., are symbolically defined within physical data stores. A semantic data model is an abstraction which defines how the stored symbols relate to the real world. Thus, the model must be a true representation of the real world.[1]

According to Klas and Schrefl (1995), the "overall goal of semantic data models is to capture more meaning of data by integrating relational concepts with more powerful abstraction concepts known from the Artificial Intelligence field. The idea is to provide high level modeling primitives as an integral part of a data model in order to facilitate the representation of real world situations".[2]

History

[edit]

The need for semantic data models was first recognized by the U.S. Air Force in the mid-1970s as a result of the Integrated Computer-Aided Manufacturing (ICAM) Program. The objective of this program was to increase manufacturing productivity through the systematic application of computer technology. The ICAM Program identified a need for better analysis and communication techniques for people involved in improving manufacturing productivity. As a result, the ICAM Program developed a series of techniques known as the IDEF (ICAM Definition) Methods which included the following:[1]

  • IDEF0 used to produce a “function model” which is a structured representation of the activities or processes within the environment or system.
  • IDEF1 used to produce an “information model” which represents the structure and semantics of information within the environment or system.
    • IDEF1X a semantic data modeling technique used to produce a graphical information model which represents the structure and semantics of information within an environment or system. Use of this standard permits the construction of semantic data models which may serve to support the management of data as a resource, the integration of information systems, and the building of computer databases.
  • IDEF2 used to produce a “dynamics model” which represents the time varying behavioral characteristics of the environment or system.

During the 1990s, the application of semantic modelling techniques resulted in the semantic data models of the second kind. An example of such is the semantic data model that is standardised as ISO 15926-2 (2002), which is further developed into the semantic modelling language Gellish (2005). The definition of the Gellish language is documented in the form of a semantic data model. Gellish itself is a semantic modelling language, that can be used to create other semantic models. Those semantic models can be stored in Gellish Databases, being semantic databases.

Applications

[edit]

A semantic data model can be used to serve many purposes. Some key objectives include:[1]

  • Planning of data resources: A preliminary data model can be used to provide an overall view of the data required to run an enterprise. The model can then be analyzed to identify and scope projects to build shared data resources.
  • Building of shareable databases: A fully developed model can be used to define an application independent view of data which can be validated by users and then transformed into a physical database design for any of the various DBMS technologies. In addition to generating databases which are consistent and shareable, development costs can be drastically reduced through data modeling.
  • Evaluation of vendor software: Since a data model actually represents the infrastructure of an organization, vendor software can be evaluated against a company’s data model in order to identify possible inconsistencies between the infrastructure implied by the software and the way the company actually does business.
  • Integration of existing databases: By defining the contents of existing databases with semantic data models, an integrated data definition can be derived. With the proper technology, the resulting conceptual schema can be used to control transaction processing in a distributed database environment. The U.S. Air Force Integrated Information Support System (I2S2) is an experimental development and demonstration of this kind of technology, applied to a heterogeneous type of DBMS environments.

See also

[edit]

References

[edit]

Further reading

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Semantic data model

View on Grokipedia
from Grokipedia
A semantic data model is a high-level in that captures the meaning, structure, and relationships of data elements in a way that closely mirrors real-world concepts, enabling more expressive and intuitive representations than traditional record-based or relational models. Unlike purely structural models, it prioritizes semantics to define entities, attributes, and interdependencies, facilitating clearer communication between domain experts and technical implementers. The concept of semantic data modeling emerged in the mid-1970s as a response to the limitations of early database models like the relational model, which often overlooked complex real-world semantics. Early examples include the Entity-Relationship (ER) model proposed by Peter Chen in 1976. One of the foundational works is the Semantic Data Model (SDM) proposed by Michael Hammer and Dennis McLeod in 1981, which introduced classes as collections of records sharing a schema, along with mechanisms for aggregation and generalization to handle inheritance and complex associations. This model advanced semantic data modeling concepts, including the Functional Data Model (FDM) by David Shipman in 1981, emphasizing functional relationships among entities, attributes, and associations. Key components of semantic data models typically include type constructors such as aggregation (grouping related objects), (subtype/supertype hierarchies), and derived elements for computed values, which reduce semantic overloading and support modularity. These features allow for explicit representation of ISA (is-a) relationships, multivalued attributes, and constraints, making the model suitable for applications requiring high abstraction levels, such as enterprise . In practice, semantic models serve as a conceptual layer that translates raw data into business terms, aiding in the creation of shared vocabularies and diagrams for . In contemporary contexts, semantic data models underpin technologies like ontologies in the and modern data platforms, including semantic layers and knowledge graphs in AI and data fabrics as of 2025, where they enable unified views of disparate data sources for and AI-driven applications. They are particularly valuable in domains such as healthcare, , and for modeling complex entities like customers or transactions while ensuring consistency and across systems.

Core Concepts

Definition and Purpose

A semantic data model is a high-level, abstract representation of data that captures the meaning (semantics) of information through entities, relationships, and attributes, independent of physical storage details. It serves as a for describing the structure and intent of an application's data environment, allowing designers to model real-world objects and their interconnections in a way that directly reflects . Unlike lower-level models focused on implementation, semantic models emphasize expressive power to represent complex semantics without ambiguity. The primary purpose of a semantic data model is to facilitate clear communication between domain experts and technical developers by providing a shared, intuitive vocabulary for data requirements. It ensures data integrity through the enforcement of business rules embedded in the model, such as constraints on relationships or attribute values, and supports scalable querying and analysis by minimizing interpretive errors across systems. By acting as a bridge from business needs to database design, it enables efficient schema evolution and serves as a formal specification tool. Core benefits include improved data reusability, as modular components like entities can be repurposed across applications; reduced , by avoiding unnecessary data duplication through semantic relationships; and enhanced , allowing integration with diverse systems via standardized meanings. These advantages stem from the model's ability to provide multiple levels, making it adaptable for both high-level planning and detailed implementation. For example, in a retail system, the entity "Customer" can be modeled with semantic attributes like "loyalty level," which implies business rules for discounts and thereby enforces consistent application of promotional logic without explicit coding in every query.

Key Components

A semantic data model fundamentally consists of entities, relationships, and attributes as its primary components, which together capture the structure and meaning of data in a domain. Entities represent real-world objects or concepts, such as "" or "," serving as the basic building blocks that encapsulate sets of similar items. Relationships define associations between entities, enabling the expression of how these objects interact, for instance, an "" link connecting a "" to a "." Attributes, in turn, describe properties of entities or relationships, such as a "Name" attribute for a "Person" entity, which can be single-valued or multivalued to reflect varying data characteristics. Semantic layers enhance these core elements by incorporating data types, domains, and integrity constraints to specify allowable values and enforce behavioral rules. Data types classify attributes into atomic forms (e.g., integers or strings) or constructed structures (e.g., sets or aggregations), ensuring precise representation of . Domains define the permissible ranges for attribute values, such as restricting a "Salary" attribute to non-negative numbers, while integrity constraints maintain consistency, including rules like uniqueness or across entities and relationships. These layers collectively imbue the model with explicit meaning, distinguishing it from purely syntactic representations. Semantic data models operate across multiple abstraction levels to bridge high-level concepts with practical implementations: the conceptual level provides an implementation-independent description of the overall domain-specific of entities and relationships; the internal level details the physical storage mechanisms; and the external level offers tailored views for end-users, focusing on relevant subsets of the data. This tiered approach supports flexibility in design and querying while preserving semantic integrity. Metadata plays a crucial role in semantic data models by adding contextual annotations that refine relationships and entities, such as cardinality constraints (e.g., one-to-many associations between "Person" and "Employment") and participation rules (e.g., mandatory involvement of a "Business" in an "Employment" relationship). These annotations, often visualized in diagrammatic notations, ensure the model accurately reflects real-world semantics and supports validation during design and use.

Historical Development

Origins in Database Design

In the early 1970s, database management systems predominantly utilized hierarchical models such as IBM's Information Management System (IMS) and network models defined by the Conference on Data Systems Languages (). These approaches enforced rigid, pointer-based structures that excelled in processing predefined paths but faltered in accommodating the intricate, many-to-many relationships and contextual meanings inherent in real-world . For instance, IMS's tree-like hierarchy often necessitated data duplication to represent non-hierarchical associations, while 's set-based navigation demanded explicit programmatic control, limiting adaptability to evolving semantic requirements. A pivotal influence emerged from the ANSI/X3/SPARC committee's 1975 interim report, which formalized the three-schema comprising external (user views), conceptual (logical structure), and internal (physical storage) levels. This framework underscored the importance of a that captures data semantics abstractly, insulated from implementation details, thereby addressing the shortcomings of earlier models by promoting and expressive modeling. Pioneering research in semantic data modeling built directly on these foundations, with David W. Shipman's 1979 introduction of the and the associated DAPLEX representing a key advancement. Shipman's model emphasized entity types, subtypes, and value-based relationships to explicitly represent and aggregation, enabling more intuitive depictions of complex domains. The overarching goal of these early efforts was to bridge the gap between human cognitive representations of information and machine-stored data, reducing the "impedance mismatch" that complicated application development and maintenance in prior systems.

Evolution and Key Milestones

In the 1980s, semantic data models advanced through integrations with the entity-relationship (ER) model originally proposed by Peter Chen in 1976, which was extended to incorporate richer semantics such as , aggregation, and subtypes to better capture real-world complexities. These extensions addressed limitations in the basic ER model by emphasizing conceptual semantics over purely structural representations. A pivotal development was the Semantic Data Model (SDM) introduced by Michael Hammer and Dennis McLeod in 1981, which supported complex objects, versioning, and multiple abstraction levels to enhance database expressiveness. The publication of "Database Description with SDM: A Semantic Database Model" in 1981 marked a key milestone, formalizing SDM as a high-level for capturing application semantics and influencing subsequent database research. Building on this, the 1990s saw standardization efforts, with semantic data models contributing to the object-oriented semantics in the (UML), released by the in 1997, particularly in class diagrams and associations that drew from semantic modeling principles. Concurrently, the emergence of XML-based models in the late 1990s enabled semantic structuring of web data, facilitating interoperability through schemas that embedded meaning beyond mere markup. Another milestone was the adoption of semantic principles in Object-Role Modeling (ORM), advanced by Terry Halpin in the 1990s, which used fact-based notation to express database constraints and rules with high semantic clarity, gaining traction in tools. Toward the end of the decade, semantic data models began linking to knowledge representation through early ontology efforts, such as in 1999, which provided a foundation for explicit semantics in distributed data, paving the way for the vision.

Modeling Techniques

Entities, Relationships, and Attributes

In semantic data modeling, entities represent the primary objects or concepts within a domain, capturing their inherent structure and semantics. In foundational models like the Semantic Data Model (SDM), entities are represented as classes, which are collections of records sharing a common schema, supporting mechanisms for components (sub-entities) and versions to handle variations over time. Entities support subtype and supertype hierarchies through generalization, where subtypes inherit attributes from supertypes, enabling more expressive modeling of domain variations—for instance, "Employee" as a subtype of "," inheriting general attributes like name and address while adding specific ones like employee ID. Relationships in semantic data models define associations between entities, incorporating rich semantics to reflect real-world interactions beyond mere connectivity. Common types include binary relationships (connecting two entities), recursive relationships (an entity relating to itself, such as an employee supervising another), and ternary relationships (involving three entities, often modeled via intermediate constructs). Semantic constructs like aggregation group entities into higher-level wholes (e.g., a convoy aggregating ships), while generalization supports inheritance hierarchies. For example, a "Manages" relationship between "Employee" and "Department" uses role names like "Manager" and "Managed Department" to clarify participation and cardinality, such as one-to-many semantics where one employee manages multiple departments. Attributes describe the properties of entities or relationships, with semantic data models distinguishing between simple, derived, and composite types to enhance meaning and . Simple attributes hold single values (e.g., an employee's date of hire), while composite attributes break into subcomponents (e.g., an address comprising street, city, and ). Derived attributes are computed from others, avoiding —such as "Salary" derived from "Base Pay" and "Bonus" components. Keys play a crucial semantic role: primary keys uniquely identify (e.g., employee ID), and foreign keys enforce relationships (e.g., department ID in an employee entity), implying and dependency semantics that prevent orphaned data. In SDM, attributes can also include derived components for computed values based on other elements. The modeling process for entities, relationships, and attributes in semantic data models involves iterative steps to identify and refine domain elements, often using adapted entity-relationship (ER) diagrams that incorporate semantic notations like ISA arcs for . First, analyze the domain to identify candidate entities as nouns, attributes as descriptive properties, and relationships as verbs, ensuring core classes precede components. Next, define hierarchies and associations, assigning keys and cardinalities to capture semantics—e.g., diagramming "Employee" inheriting from "" via and linking to "Department" via "Manages." Finally, refine by validating dependencies, deriving attributes where appropriate, and normalizing to eliminate redundancies while preserving meaning, resulting in a that supports query optimization and extensibility.

Semantics, Constraints, and Formalisms

In semantic data models, semantics are defined through both explicit annotations, such as direct specifications of object meanings via attributes and relationships, and implicit inferences derived from the model's structural arrangements, like hierarchical inclusions that imply broader categorizations. This dual approach plays a crucial role in disambiguating data interpretation by clarifying the intended meaning of entities beyond mere syntax, ensuring that data elements are understood in context-specific ways, such as distinguishing a "bank" as a versus a riverbank through relational ties. Constraints in semantic data models enforce and business rules, categorized into static, dynamic, and semantic types. Static constraints govern value ranges and structural properties at a fixed point, such as limiting an attribute like "age" to non-negative integers. Dynamic constraints manage transitions over time, including update propagation rules that maintain consistency during data modifications, like ensuring when altering relationships. Semantic constraints capture domain-specific invariants, such as business rules that integrate meaning to prevent invalid interpretations. Formalisms in semantic data models typically rely on graph-based structures and set-theoretic definitions to represent entities, relationships, and type constructors like aggregation and , enabling of properties. Validation techniques involve consistency checking to ensure the model admits a non-empty interpretation and satisfies all axioms, often using algorithms to detect issues like cycles in generalization hierarchies.

Comparisons with Other Models

Versus Relational and Entity-Relationship Models

Semantic data models differ from relational models primarily in their emphasis on capturing inherent meaning and business rules, which are absent in the flat, tabular structures of relational databases. While relational models, as introduced by Codd in 1970, focus on data normalization—such as achieving (3NF) to minimize redundancy and ensure —they treat data primarily as atomic values in tables without built-in support for semantic constructs like aggregation or derivation rules. In contrast, semantic models incorporate layers of interpretation, such as explicit representations of real-world concepts and constraints. Research suggests that this semantic richness leads to improved modeling performance and comprehension. Compared to the entity-relationship (ER) model proposed by Chen in , which provides a structural framework using entities (represented as rectangles), relationships (diamonds), and attributes (ovals) to diagram database schemas, semantic data models extend this foundation with advanced features for richer expression. ER models capture basic semantics through entity interconnections but lack native support for subtypes, inheritance hierarchies, derived attributes, or complex constraints, limiting their ability to model intricate real-world scenarios without extensions. Semantic models address these gaps by incorporating mechanisms like supertype/subtype relationships and formal integrity rules, allowing for more precise depiction of hierarchies and dynamic properties that ER diagrams handle only superficially. One key advantage of semantic data models is their superior handling of complex hierarchies and , which facilitates and in , reducing errors in representing overlapping or specialized types—areas where both relational and ER models require additional normalization or diagrammatic workarounds. However, this higher level of can complicate mapping to physical schemas, as semantic constructs must be translated into relational tables, potentially introducing overhead or decisions like join strategies. For example, semantic subtypes can be mapped using table-per-hierarchy strategies with a type discriminator or table-per-type approaches to preserve while maintaining relational integrity.

Versus Ontologies and Graph-Based Models

Semantic data models, such as the Semantic Data Model (SDM) proposed by and McLeod, are primarily database-centric frameworks designed to capture the meaning of data through structured entities, relationships, and attributes for application-specific schema design. In contrast, ontologies, formalized in languages like OWL, operate at a higher semantic level to represent generic independently of specific implementations, emphasizing reusability across applications. While semantic data models focus on articulating organizational data structures with less emphasis on logical inference, ontologies prioritize axiomatic definitions that enable and knowledge integration in heterogeneous environments. Graph-based models, including RDF for applications and property graphs in modern databases, represent data as nodes and edges to facilitate flexible traversals and queries over interconnected information. Semantic data models, however, prioritize predefined semantics through type constructors like generalization (ISA hierarchies) and aggregation to ensure enterprise-wide consistency and , rather than the schema-optional flexibility of pure graph structures. For instance, RDF triples (subject-predicate-object) support open-world assumptions for , whereas semantic data models enforce closed-world constraints akin to traditional databases for validation during schema evolution. Both semantic data models and graph-based approaches overlap in their use of graph-like representations for knowledge encoding, such as entities connected by relationships, which aids in modeling complex interdependencies. However, semantic data models integrate explicit constraints and derived components for , distinguishing them from the traversal-focused, often schema-less nature of RDF or property graphs that may lead to without additional formalisms.

Modern Applications

In Business Intelligence and Data Warehousing

In business intelligence (BI), semantic data models serve as a semantic layer that abstracts underlying physical data sources into intuitive business terminology, enabling users to query and analyze data using familiar concepts such as "Customer Lifetime Value" or "Revenue by Region" without needing to understand complex database schemas. In tools like Microsoft Power BI and Tableau, this layer translates raw data into a unified model that supports ad-hoc reporting and visualization, fostering a consistent view across disparate sources like transactional databases and cloud storage. For instance, Power BI's semantic model leverages relationships between tables to enforce business rules, ensuring that metrics like sales totals are calculated accurately regardless of the underlying data structure. Within data warehousing, semantic data models enhance traditional techniques, such as s, by incorporating semantics to define clear relationships between fact and dimension tables, thereby improving query efficiency and . In a , fact tables hold quantitative measures (e.g., order quantities), while dimension tables provide contextual attributes (e.g., product categories); semantic extensions add constraints that govern measure aggregation, such as ensuring additive rules for sums across time periods or non-additive handling for ratios like conversion rates. This integration allows warehouses to maintain a logical over physical storage, supporting scalable in environments like Fabric, where semantic models optimize joins and filters for large datasets. The adoption of semantic data models in BI and warehousing yields significant benefits, including empowered self-service analytics where business users can explore independently without IT intervention, and reduced query complexity through pre-defined metrics that minimize errors in reporting. Tools like GoodData's , which saw expanded integration post-2021, exemplify this by providing governed access to metrics across BI applications, enabling consistent insights for thousands of users while cutting development time for new reports. Overall, these models promote democratization, with studies indicating that projects can be completed up to four times faster and costs cut by 50% in governed environments compared to traditional approaches. However, implementing semantic layers in large-scale data warehouses introduces challenges, particularly performance overhead from real-time translations and complex rule evaluations, which can increase query latency on petabyte-scale datasets. Solutions include in-memory caching to store frequently accessed semantic definitions and query optimization techniques, such as materializing aggregates in the warehouse layer, which mitigate these issues and maintain sub-second response times even under high concurrency.

In Semantic Web, Knowledge Graphs, and AI

Semantic data models form the cornerstone of the , providing a structured framework for representing and interlinking data across the web to enable machine-readable semantics and interoperability. The (RDF), standardized by the (W3C) in 2004, serves as a foundational model for encoding data as triples consisting of subjects, predicates, and objects, using Uniform Resource Identifiers (URIs) to uniquely identify entities and relationships. This allows for the creation of , where disparate datasets can be merged seamlessly based on shared vocabularies, facilitating automated reasoning and discovery. Complementing RDF, the (OWL), also a W3C recommendation from 2004, extends these capabilities by defining ontologies that specify classes, properties, and inference rules, enabling richer semantic expressions such as subclass relationships and cardinality constraints. A prominent example is Schema.org, a collaborative vocabulary developed by major search engines since 2011, which provides extensible schemas for embedding structured data in web pages using , Microdata, or formats to enhance search engine understanding and rich result displays. In knowledge graphs, semantic data models underpin the organization of vast, interconnected entity networks, supporting advanced querying and inference for real-world applications. Google's , launched in 2012, leverages semantic modeling to connect entities like people, places, and concepts through typed relationships, drawing from sources such as Freebase and to deliver contextually relevant search results via infoboxes and knowledge panels. This approach enables entity resolution—identifying and linking equivalent entities across datasets—and powers recommendation systems by inferring implicit connections, such as related products or historical events. In enterprise settings, tools like extend graph databases with semantic overlays, where RDF/OWL ontologies are integrated to add explicit meaning to nodes and edges, facilitating domain-specific queries for tasks like fraud detection and . For instance, semantic layers in allow for hybrid traversals that combine structural graph patterns with ontological constraints, improving accuracy in entity disambiguation and . The integration of semantic data models with , particularly large language models (LLMs) and AI agents, enhances reliability by grounding outputs in verifiable, context-rich structures. As of 2025, platforms like Appsmith emphasize for AI data grounding, using graph databases such as to create knowledge layers that provide business-specific semantics, enabling agents to query and reason over enterprise data without relying solely on probabilistic embeddings. This semantic context reduces hallucinations—fabricated or inaccurate responses in LLMs—by anchoring generations to explicit entity relationships and constraints, with studies showing significant error reduction in domain-specific tasks through verified semantic caches. For example, in retrieval-augmented generation (RAG) pipelines, semantic models supply structured metadata to LLMs, allowing for more precise and response synthesis in applications like chatbots. Post-2023 advancements have introduced hybrid semantic models in vector databases, merging ontological structures with embedding-based similarity search to optimize RAG systems for complex queries. These hybrid approaches, such as those in HelixDB, combine graph-based semantics for relational with vector representations for fuzzy matching, enabling systems to retrieve both exact matches and semantically similar content in a single pass. In practice, this fusion supports enhanced RAG by incorporating sparse vector techniques alongside dense embeddings, improving retrieval precision in knowledge-intensive tasks like legal document , where structural constraints from semantics mitigate embedding drift. Such models, often built on frameworks like or with vector extensions, demonstrate scalable performance in multimodal RAG, handling text, images, and graphs while preserving semantic integrity for AI-driven decision-making. As of September 2025, GoodData's acquisition of Understand Labs further advances semantic integration for AI data in enterprise analytics.

References

  1. https://sites.cc.gatech.edu/computing/Database/readinggroup/articles/p201-hull.pdf
  2. https://www-cs.stanford.edu/people/chrismre/cs345/rl/datamodel.pdf
  3. https://www.ontoforce.com/blog/gartner-semantic-technologies-take-center-stage-in-2025
  4. https://dl.acm.org/doi/10.1145/509252.509264
  5. https://cs.franklin.edu/~crawforl/.c281/ppts_text/ch02text.pdf
  6. https://people.cs.umass.edu/~yanlei/courses/CS691LL-f06/papers/SH05.pdf
  7. https://www.cs.ubc.ca/~rap/teaching/504/2023W2/slides/whatgoesaround.pdf
  8. https://nvlpubs.nist.gov/nistpubs/Legacy/IR/nbsir85-3173.pdf
  9. https://dl.acm.org/doi/10.1145/45072.45073
  10. https://www.sciencedirect.com/science/article/pii/0169023X85900278
  11. https://dl.acm.org/doi/10.1145/319587.319588
  12. https://www.orm.net/pdf/EncDBS.pdf
  13. https://docs.teradata.com/r/Enterprise_IntelliFlex_VMware/Database-Design/Semantic-Data-Modeling/Entities-Relationships-and-Attributes
  14. https://redirect.cs.umbc.edu/courses/graduate/691/fall17/01/papers/DescriptionLogicHandbook.pdf
  15. https://static.aminer.org/pdf/PDF/000/002/576/semantic_data_models.pdf
  16. https://www.researchgate.net/publication/220373867_A_Meta-Analysis_Comparing_Relational_and_Semantic_Models
  17. https://dl.acm.org/doi/10.1145/320434.320440
  18. https://www.researchgate.net/publication/235625479_Semantic_Data_Models
  19. https://dl.acm.org/doi/10.1145/111197.111200
  20. https://dl.acm.org/doi/10.1145/38713.38732
  21. https://agiledata.org/essays/mappingobjects.html
  22. https://tomgruber.org/writing/definition-of-ontology.pdf
  23. https://link.springer.com/article/10.1007/s10115-021-01558-4
  24. https://link.springer.com/article/10.1007/s41019-016-0029-6
  25. https://dl.acm.org/doi/10.1145/637411.637413
  26. https://learn.microsoft.com/en-us/power-bi/guidance/star-schema
  27. https://www.tableau.com/blog/your-semantic-data-model-secret-trusted-agentic-analytics
  28. https://www.datacamp.com/blog/semantic-layer
  29. https://learn.microsoft.com/en-us/fabric/data-warehouse/dimensional-modeling-overview
  30. https://www.sciencedirect.com/science/article/abs/pii/S0169023X11000310
  31. https://www.getdbt.com/blog/semantic-layer-introduction
  32. https://www.gooddata.com/blog/what-is-a-semantic-layer/
  33. https://www.atscale.com/glossary/semantic-layer/
  34. https://www.databricks.com/glossary/semantic-layer
  35. https://airbyte.com/blog/the-rise-of-the-semantic-layer-metrics-on-the-fly
  36. https://www.w3.org/news/2004/rdf-and-owl-are-w3c-recommendations/
  37. https://www.w3.org/OWL/
  38. https://schema.org/
  39. https://cacm.acm.org/practice/schema-org/
  40. https://blog.google/products/search/introducing-knowledge-graph-things-not/
  41. https://neo4j.com/blog/knowledge-graph/from-graph-to-knowledge-graph-how-a-graph-becomes-a-knowledge-graph/
  42. https://go.neo4j.com/rs/710-RRC-335/images/Building-Knowledge-Graphs-Practitioner%27s-Guide-OReilly-book.pdf
  43. https://neo4j.com/blog/developer/knowledge-graph-structured-semantic-search/
  44. https://www.appsmith.com/blog/semantic-data-model-blind-spot-ai-agents
  45. https://www.dataart.com/blog/prevent-ai-hallucinations-with-semantic-data-models
  46. https://aws.amazon.com/blogs/machine-learning/reducing-hallucinations-in-llm-agents-with-a-verified-semantic-cache-using-amazon-bedrock-knowledge-bases/
  47. https://www.nature.com/articles/s41586-024-07421-0
  48. https://arxiv.org/html/2510.10123v1
  49. https://www.researchgate.net/publication/390326215_Hybrid_Retrieval-Augmented_Generation_RAG_Systems_with_Embedding_Vector_Databases
  50. https://memgraph.com/blog/why-hybridrag
  51. https://milvus.io/blog/milvus-in-2023-unprecedented-vector-database-amidst-tech-buzz.md
  52. https://www.gooddata.com/press-releases/gooddata-acquires-understand-labs-accelerates-ai-data-storytelling-and-agentic-vision/
Add your contribution
Related Hubs
Contribute something
User Avatar
No comments yet.