188.399 Introduction to Semantic Systems

188.399 Introduction to Semantic Systems

• This course, taught by Max Tiessler, covers semantic systems
• The document is a handbook/study guide for the course
• The handbook includes sections on Introduction, Ontology, Types and Categories (lightweight and heavyweight), Ontology Mechanics, Description Logics, Building Ontologies, Semantic Web, RDF, RDFS, OWL, SPARQL, SHACL, Knowledge Graph Creation, RDF Storage, and Extras.
• The handbook is for the Winter Semester 2025
• Material is available on GitHub at: https://github.com/mtiessler/188.399-Introduction-To-Semantic-Systems-Summary-TUW/blob/main/ISS_Summary.pdf

2 What is an Ontology?

• Ontology is the systematic study of existence
• According to the Merriam-Webster Dictionary:
  • A branch of metaphysics concerned with the nature and relations of being
  • A particular theory about the nature of being and kinds of existents.
• Formal: Machine-readable, explicitly defined.
• Explicit: Concepts, properties, functions and axioms are clearly defined.
• Shared: The ontology is agreed upon by the community.
• Conceptualization: Provides an abstract model of phenomena in the world.

2.2 Types and Categories

• Lightweight ontologies are simple frameworks for knowledge organization.
  • Controlled Vocabulary: A finite list of terms (e.g., library catalog)
  • Glossary: A controlled vocabulary with natural language definitions
  • Thesauri: A controlled vocabulary connecting concepts via various relations (equivalency, hierarchies, associations).
• Heavyweight ontologies rigorously define and organize knowledge.
  • Rigorous Definitions: Concepts and relationships are precisely defined.
  • Focus on Logic: Formal rules drive deductions and inferences.
  • Formal Specifications: Detailed, consistent definitions to eliminate potential conflicts.

2.3 Ontology Mechanics

• Ontology mechanics involve the structure and logical rules governing relations and organization.
• These include establishing domain and range relations, defining and maintaining consistency, and enforcing constraints like disjointness or cardinality.

2.4 Description Logics

• Description Logics (DL's) are formal knowledge representation languages used in ontologies.
• They are a subset of First-Order Logic used for describing concepts.
• Key features of DLs
  • Syntax: Rules for valid expressions.
  • Semantics: Meaning of those expressions.
  • Calculus: Methods to determine the meaning.

2.5 How to Build an Ontology

• A step-by-step, somewhat iterative process
  • 1. Determine Scope: Define the ontology's purpose and planned extensions.
  • 2. Consider Reuse: Leverage existing, validated ontologies in similar domains.
  • 3. Enumerate Terms: List all necessary concepts and terms.
  • 4. Define Classes and Taxonomy: Group terms into classes organized hierarchically.
  • 5. Define Properties: Describe attributes of classes and the relationships between them.
  • 6. Define Constraints: Refine the ontology by adding logical rules.
  • 7. Create Instances: Represent examples of concepts (e.g., individual entities)
  • 8. Check Anomalies: Validate through the ontology's formal semantics

3 Semantic Web, RDF, RDFS, OWL

• The Semantic Web is a machine-readable extension of the current web

• Its goal is to enable computers to interpret knowledge and coordinate tasks.

• The Semantic Web is based on Linked Data principle

  • links across different resources rather than just documents.
  • the links are qualified to define relationships.

• Two main characteristics of the web : The Web of Documents and Web of Data (Semantic Web) which focus on human-readable pages vs. machine-readable data.

3.2 RDF

• RDF (Resource Description Framework) is a standard framework for the Semantic Web.
• RDF statements are triples, with a subject, predicate, and object.

3.3 Advanced RDF

• Advanced RDF expands core RDF principles into a more robust model for intricate systems
• Includes features like structured values, unnamed resources and metadata for statements.

3.4 RDFS

• RDFS (RDF Schema) is an extension of RDF enabling semantic structures and interoperability
• Defines "types" of objects about which assertions can be made.
• Defines properties used as predicates in the assertions.

3.5 OWL (Web Ontology Language)

• OWL extends RDFS, providing a fully-fledged knowledge representation language with logical constructs and formal semantics.
• Allows more complex descriptions and supports inferencing.
• OWL expands RDFS with enhanced logic, expressiveness, and capabilities.

4 SPARQL

• SPARQL (SPARQL Protocol and RDF Query Language) is a standard query language for RDF data.
• Supports querying and manipulating RDF data.
• Has distinct query forms (SELECT, ASK, DESCRIBE, and CONSTRUCT)
• Supports various patterns for querying more complex information and handling optional patterns.
• Includes solution modifiers (filtering, ordering, aggregation)
• SPARQL 1.1 introduced data manipulation functions like INSERT, DELETE and DROP.

5 SHACL

• SHACL (Shapes Constraint Language) is used to validate and describe constraints on RDF graphs.
• Defines shapes to capture specific structures and relationships in the RDF Data,
• Includes Node Shapes (constraints on a particular node/concept) and Property Shapes (constraints on properties/relationships).

6 Knowledge Graph Creation

• Knowledge graphs link data from diverse sources to provide a comprehensive, semantically rich representation.
• Different methods exist for knowledge graph development, including
  • Template-based Generation: Use existing web infrastructures (e.g. JSON-LD) to transform data into RDF.
  • Mapping and Transformation: Transfer data from various sources (e.g., databases, spreadsheets, unstructured text) into RDF, leveraging frameworks like RML (RDF Mapping Language).
• These approaches allow for integration with existing knowledge graphs (e.g., Wikidata and DBpedia).

6.2 Tabular Data to RDF

• Techniques on handling and restructuring tabular data
• OpenRefine: A spreadsheet-like tool for efficiently cleaning, transforming, and reconciling messy data.
• Reconciliation services: to match names in tables with canonical references from authoritative knowledge sources (e.g., Wikidata).

6.3 Relational Data to RDF

• Standard W3C approaches for mapping relational databases to RDF (resource, predicate, object)
• Direct mapping: Simplifies the process by automatically generating triples.
• Flexible R2RML: Allows full customization of RDF mappings, more complex scenarios (joins).

6.4 Structured Data to RDF

• Techniques for knowledge graph development on structured data formats (JSON, XML, CSV files)
• RML: RML (RDF Mapping Language) is a generalization of R2RML capable of mapping and interpreting non-tabular data as RDF triples.

6.5 Text to RDF

• Extracting structured information from unstructured text resources like documents and articles.
• Key technique (Annotation): Assigning explicit, structured information (annotations) to text segments (creating statements).
• Tool examples: DBpedia Spotlight and PermID for identifying and linking concepts from free-text data.

6.6 RDF Storage

• Storage approaches to RDF knowledge graphs (knowledge graph databases or graph databases).
• Categorization: Centralized (one machine) vs. Distributed (across a cluster of machines) with example systems and techniques.
• Strategies: Statement Tables, Property Tables, Index Permutations, and Vertical Partitioning.

7 Extras

• Historical Context: Knowledge graph development (evolution, and goals).
• Industry Impact: Knowledge graphs play a role in e-commerce, industry management, and many other fields.
• Neurosymbolic AI: Combining neural networks with symbolic reasoning, with knowledge graphs supporting the integration of data and reasoning.
• Data Fabrics and Semantic Data Lakes: Methods for creating a unified repository of data (data lakes).
• Provenance and Explainability: Knowledge graphs explain processes and steps based on the inputs and steps for decision-making.
• Use Cases & Directions: Different areas of application and their future directions (healthcare, economy, media, transportation).

This study guide for '188.399 Introduction to Semantic Systems' offers a comprehensive overview of semantic systems, including ontology, description logics, and semantic web technologies. It is specifically designed for the Winter Semester 2025 and serves as a valuable resource for students engaged in this advanced topic.