You are on page 1of 25

Universidade de Coimbra

Faculdade de Ciências e Tecnologia


Departamento de Engenharia Informática

Knowledge-Based Systems
Masters in Informatics Engineering

Course Overview

Hugo Gonçalo Oliveira


hroliv@dei.uc.pt
2020-2021

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 1 / 25


Overview

1 Introduction

2 The Knowledge Base

3 The Semantic Web

4 Knowledge Acquistion

5 Natural Language Processing

6 Semantic Search

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 2 / 25


Introduction

What are Knowledge-Based Systems?

Knowledge-Based System
A computational system that draws upon the knowledge of human experts
on a particular domain, captured in a Knowledge Base, to solve problems
that typically require humans.

Knowledge
Base
Domain
User
Knowledge
Expert
Interface
Engineer
Inference
User
Engine

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 3 / 25


Introduction

Components of a Knowledge-Based System

Knowledge Base (KB)


Facts and rules that represent a domain
Often created by a domain expert or with the support of one.
Represented in a Knowledge representation Language.

Inference Engine
Algorithms that, following a given query / interaction / perception,
use the facts and rules of the KB for arriving at a decision.

User Interface
Means for interacting with the end user.

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 4 / 25


The Knowledge Base

Knowledge Base

Knowledge Representation
Dedicated to representing information about the world / a domain in a
form that a computer system can utilize to solve complex tasks.
Facts + Rules

Knowledge Representation Formalisms


Often based on logic
Propositional logic
First-order logic

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 5 / 25


The Knowledge Base

Knowledge Base

Knowledge Representation Languages


Formal specifications for representing things and their properties
Document Person
isA isA
HasAuthor
Paper#481 John Smith

Text:
Human language is probably the most natural way of encoding, transmitting
and reasoning about knowledge
Most knowledge repositories (e.g., libraries, the Web) are available
exclusively in written form!

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 6 / 25


The Semantic Web

The Semantic Web Vision

The Semantic Web (Tim Berners-Lee, 2001)


“An extension of the Web in which information
is given well-defined meaning, better enabling
computers and people to work in cooperation”

+ Semantic
Web
Layer
=

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 7 / 25


The Semantic Web

Semantic Web Languages

Represent the meaning of each resource (on the Web)


Identify resources with URIs
Use RDF for stating their properties (types, labels, user-defined) and
restrictions (domain, range, cardinality, ...)
pub -481 type Publication
person -7152 type Person
org -294 type Organization
course - m366 type Course
country -76 type Country
book -34571 type Book
pub -481 label " Pub1 "
person -7152 label " Mathieu "
org -294 label " The Open University "
country -76 label " Belgium "
book -34571 label " Mechatronics "
pub -481 author person -7152
person -7152 workFor org -294
org -294 offers course - m366
course - m366 availableIn country -76
course - m366 setBook book -34571

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 8 / 25


The Semantic Web

Linked Data
Reusable vocabularies available for representing knowledge in RDF
Several RDF datasets linked in the Web of Linked Data
Useful for better understanding world knowledge

https://lod-cloud.net/
hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 9 / 25
The Semantic Web

Popular Knowledge Bases

Though developed with the Web on mind, Semantic Web languages


can be used for any kind of Knowledge Representation
Several large Knowledge Bases are represented in RDF!

...
hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 10 / 25
The Semantic Web

Linked Data
Example

athlete
hypernym
hasJob
soccer player
gloss
nationality suffersFrom an athlete who plays soccer

WordNet
birthDate name

bornIn GHD
1987-06-24 Lionel Andrés Messi
isA
DBPedia
cityIn capital population hypopituitarism

Rosário Buenos Aires 44494502


isA
coordinates
pituitary gland disease
-34.61315,-58.37723
Disease
GeoNames

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 11 / 25


The Semantic Web

Knowledge Reasoning

SPARQL Protocol and RDF Query Language


Additional rules can be written for reasoning

Born: 1987
Nation: Argentina
Club: FC Barcelona
SPARQL query Height: 1.7 m
Born: 1985
Nation: Portugal
Club: Juventus FC
Height: 1.87 m
Born: 1985
Nation: Croatia
Club: Real Madrid CF
Height: 1.72 m

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 12 / 25


Knowledge Acquistion

Limitations of Knowledge Based Systems

Building a suitable knowledge base can be very challenging...


Time-consuming
Trade-off between capturing all necessary expert knowledge and still
have a consensual representation of the problem domain
Unlikely to code all the knowledge
Knowledge can change

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 13 / 25


Knowledge Acquistion

Knowledge Acquistion

Alternative approaches for creating a large knowledge base


Collaboratively created by human amateur volunteers
ConceptNet, originally based on common sense knowledge collected
from human volunteers
Automatically from structured sources
DBPedia, knowledge base extracted from Wikipedia
Automatically from unstructured sources (e.g., text)
Natural Language Processing, Information Extraction,
Ontology Learning

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 14 / 25


Natural Language Processing

Natural Language Processing


Much knowledge simply represented in human language!
Not directly accessible by computers...
Many challenging phenomena, such as:
Unstructured, different languages
Linguistic phenomena: ambiguity, vagueness, irony, ...

Natural Language Processing


Enabling machines to understand human language and use it to
communicate with humans.
hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 15 / 25
Natural Language Processing

Information Extraction for Knowledge Base Creation

Apple acquired Siri, a company working on voice recognition.

acquired

isA company

worksOn voice recognition


hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 16 / 25
Natural Language Processing

Distributional Semantics
Distributional Hypothesis
Words that occur in the same contexts tend to convey similar meanings.
animal
keyboard device

dog mouse laptop


cat computer
tiger bird aircraft
lion man vehicle
woman tank
king
queen car
beverage
bike
wine
soda beer
whisky

Words represented as vectors in a n-dimensional space


Machine Learning-friendly!
Linguistic regularities
similarity(cat, dog) = cos(cat,~ dog ~ )
similarity (cat, dog ) > similarity (cat, beer )
~ ≈ king
queen ~ − man ~ + woman~
~
Knowledge Acquisition ? vehicle ~
≈ beverage ~ + car
− wine ~
hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 17 / 25
Natural Language Processing

Traditional Information Retrieval

"University of Coimbra"

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 18 / 25


Semantic Search

Semantic Search

Knowledge
Base

query

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 19 / 25


Semantic Search

Semantic Search
Keyword-based

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 20 / 25


Semantic Search

Semantic Search
Question Answering & Natural Language Interfaces for Knowledge Bases

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 21 / 25


Summary

Syllabus

Knowledge Representation
The Semantic Web
Semantic Web technologies
Representation languages: RDF, RDFS, OWL
Ontologies
Querying & Reasoning: SPARQL
Knowledge Acquisition
Automatic creation of knowledge bases
Natural Language Processing
Information Extraction
Distributional Semantics
Knowledge Retrieval
Information Retrieval
Semantic Search
Natural Language Interfaces for Knowledge Bases

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 22 / 25


Summary

Main Bibliography

Jurafsky, D. and Martin, J. H. (2009).


Speech and Language Processing: An Introduction to
Natural Language Processing, Computational Linguistics
and Speech Recognition.
Prentice Hall series in Artificial Intelligence. Prentice
Hall, Pearson Education International, Englewood Cliffs,
NJ, 2nd edition.

Maynard, D., Bontcheva, K., and Augenstein, I. (2016).


[Russell and Norvig, 2010] [Yu, 2015]
Natural Language Processing for the Semantic Web.
Morgan & Claypool Publishers.

Russell, S. J. and Norvig, P. (2010).


Artificial Intelligence (A Modern Approach).
Prentice Hall.
Yu, L. (2015).
A Developer’s Guide to the Semantic Web.
Springer, 2nd edition.
[Maynard et al., 2016] [Jurafsky and Martin, 2009]

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 23 / 25


Summary

Questions?

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 24 / 25


Summary

hroliv@dei.uc.pt (DEI-FCTUC) Knowledge-Based Systems 2020/2021 25 / 25

You might also like