Knowledge base definitions

Risultati immagini per knowlege

Contents

1 What is a knowledge base
2 What is language profile
3 What is a sharable knowledge
4 What is a semantic web agent
5 The knowledge base life cycle
6 How knowledge base management is implemented in LinkedData.Center

What is a knowledge base

In LinkedData.Center a Knowledge Base is a semantic system where information is described as a set of statements according with the W3C standard Resource Description Framework (RDF).

There are two disjoint set of statements: *TBox* and *ABox*. Together they make up the whole *knowledge base*:

TBox statements describe a system in terms of controlled vocabularies, for example, a set of classes and properties such as: **All Students are Persons** or **There are two types of Persons: Students and Teachers**.
ABox statements describe facts about things, using vocabularies defined in TBox such as **John is a Student** or **Mary has blue eyes**.

In other words, TBox statements sometimes associate with object-oriented classes and ABox statements associate with instances of those classes. TBox statements tend to be more permanent within a knowledge base. and are often grouped in "ontologies" that describe a specific knowledge domain (e.g. business entities, people, goods, friendship, offering, geocoding, etc, etc)

In contrast, ABox statements are much more dynamic in nature. A special type of ABox statements are those that are inferred through an automatic reasoning process, that is the evaluation of rules and/or axioms explicitly asserted or derived from TBox.

What is language profile

In order to let computers do work for us, they must understand data. Not only grammar and syntax (like in EDI) but the real meaning of things. When this happens, machines can recognize, manage data and make decisions even when their provenance and quality is not totally known and data are incomplete. To enable this, we need to share knowledge using a language that is understandable both for humans and machines.

A language profile is a trimmed down version of a set of ontologies that trades some expressive power for the efficiency of automatic reasoning.

LinkedData.Center recognizes the KEES (Knowledge Exchange Engine Schema) language profile TBox statements and uses them to drive automated data ingestion.

What is a sharable knowledge

LinkedData.Center introduces the concept of sharable knowledge fragment as a set of ABox statements associated to KEES language profile (TBox) that describe the knowledge base content in terms of:

ingestion policies: that state how and when to populate the knowledge base from linked data ad rdf datasources.
inference rules: that describe how to generate new facts from RDF statements.
included knowledge fragments: that are external linked data sources containing other knowledge fragments to import

A knowledge fragment is usually a linked data web resource. All knowledge fragments in a knowledge base are referred as the Knowledge Base Configuration.

What is a semantic web agent

A semantic web agent denotes the piece of software that possesses the properties of autonomy, social ability, reactivity, proactivity, temporal continuity, and goal-orientedness.

Multi-agent system consists of a number of agents which are capable of interacting with each other. In these systems, the agents are capable to cooperate, coordinate, and negotiate with each other.

LinkedData.Center provides a multi-agent system to automate most of the typical knowledge base managements tasks:

knowledge sharing agents: that manage knowledge fragments and the knowledge base structure;
data learning agents: that populate and maintains TBox and Abox statements;
reasoners: that manage axioms and rules to infer new data from existing facts;

The knowledge base life cycle

LinkedData.Center adopts a functional model based on a sequence of four temporal phases called "windows“:

an initial knowledge base initialization phase (booting window) when all knowledge fragments are imported in a knowledge base by knowledge sharing agents;
a time slot for the population of a Knowledge Base (learning window) when ingestion policies are evaluated and Linked Data (ABox and/or TBox) statements are indexed in aknowledge base by knowledge learning agents;
a time slot for the data inference (reasoning window) when inference rules are evaluated and new facts inferred by reasoners;
a time slot for querying the knowledge base (teaching window).

A LinkedData.Center ingestion activity consists in the sequential execution of one or more of the first three windows. Typically the booting window is requested after a changing in knowledge base configuration, while learning and reasoning windows are requested on dataset changes .

During an ingestion activity, the Knowledge base can be in an inconsistent status. In this period LinkedData.Center returns 503 http status (Service Unavailable) to all requested queries; anyway you can bypass this "guard" in order to have a look inside your knwolege base while it is forming.

You request an ingestion activity through the Data ingestion activity API.

To query the knowledge base use LinkedData.Center provides a SPARQL query end point fully compliants with W3C standards.

A SPARQL 1.1. update interface is also available t

How knowledge base management is implemented in LinkedData.Center

LinkedData.Center implements a knowledge base according W3C Semantic web best practices and standards providing:

an RDF graph database instance to store rdf statements;
a set of RESTful services (APIs) to manage the LinkedData.Center semantic web agents for data ingestions, reasoning and knowledge base configuration;
a SPARQL Query endpoint to query the knowledge base according with SPARQL protocol query operations;
a SPARQL Update endpoint o update, manually or programmatically, the knowledge base according with SPARQL protocol update operations (*);

(*) Note that manually changing the knowledge base through SPARQL update could expose you to the risks to introduce inconsistent data and to reduce the efficiency of LinkedData.Center cache.

The RDF graph database is partitioned in named graphs and in particular:

a set of named graphs that contain linked data (ABox and TBox);
a special graph containing all inferred facts;
a set of graphs that contain ingestion activity reports.

graph type	content type	graph name	life cycle
LinkedDataGraph	linked data statements	same as the source URL	These graphs are loaded during learning window. They are reloaded on data source changes.
InferredKnowledgeGraph	Inferred facts	urn:kees:inferences	This graph is created during reasoning window every time a linked data source changes

All graphs contain RDF data triples together with their provenance metadata. The default graph contains the union of all named graph contents.

License