An OWL Framework For Rule-Based Recognition of Places in Italian Non-Structured Text

Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
An OWL framework for rule-based recognition of

places in Italian non-structured text
Domenico Cantone Andrea Fornaia
Marianna Nicolosi-Asmundo Daniele Francesco Santamaria
Emiliano Tramontana
Department of Mathematics and Computer Science, University of Catania
KDWEB 2016 - Cagliari, September 9, 2016

Work supported by the project PRIME - Piattaforma di Reasoning Integrata,
Multimediale, Esperta - PON FESR Sicilia 2007/2013 and by the FIR project
COMPACT: Computazione affidabile su testi firmati, code: D84C46.
Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana
1 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Contents
Introduction
Motivation
Approach Used
ASM Algorithm
Pipe and Filter
Multi-Agent Model
Phase 1: Preliminaries
Phase 2: Rules
Phase 3: Further filtering
OWL Framework
Semantic Web: Overview

The Ontologies
The Ontologies: Places
Modelling
The Ontologies:
Estimation Modelling
The Ontologies:
Algorithms Modelling
Examples
Future Works
Bibliography
2 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Motivation
Approach Used
Motivation
Recognition of location names inside non-structured text has
several practical applications:
investigative field to analyse interception transcriptions.
social media context to reveal visited places to target
advertisements.
Different approaches were studied in the literature:

Automatic learning, maximum entropy, conditional random
fields...
Linked data and ontologies have been used to address the

question only in the last decade.
Existing approaches mainly focus on English; they are hard to
generalize for other languages.
3 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Motivation
Approach Used
Approach Used
Recognition of location names belonging to the Italian

country appearing in non-structured Italian text.
We first apply a Rule-based algorithm to detect location
names.
Then we use several semantic web tools to store data and
reason on them even in case of ambiguities:
open geographic datasets.
OWL ontologies.
4 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:
Filter Multi-Agent Model

Preliminaries
Rules
Further filtering
Pipe and Filter Multi-Agent Model

Agents implementing grammar rules to spot a location name.
Il vertice che si terra oggi a Bruxelles (The meeting taking
place today in Bruxelles)
Three phases, connected in a pipe & filter style.

Preliminary results: F1 score is 0.67 (max).
Rule 1
sentence
splitting
Rule 2
Filter0
capital
letters
Filter1
nonplaces
Filter2
verbs
Rule 3
Phase 1
Phase 2
Phase 3
5 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 1: Preliminaries
Longer texts, e.g. a novel, are split into sentences.

A sentence is tokenised based on a lexicon.
Articles. Italian definite and indefinite articles.
Prepositions. Both simple and composite ones, excluding con
(with).
Verbs. Verbs that may be related to places.
Descriptors. Adverbs that may be related to places.
Non-places. Additional customisable list of words already
known as false positives.
6 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 2: Rules
Three rules, based on Italian sentence patterns possibly
related to a location.
Each agent implements a rule, by using an automaton.
It works at the token level.
The accepting state of the automaton identifies a candidate
location.
Each rule is independent from the others.
Rule 1: Da Roma.
Rule 2: Vicino a Roma.
Rule 3: Andando a Roma.
7 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 2: Rule 1 (Da Roma)
8 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 2: Rule 2 & Rule 3
9 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 2: Rule 1, 2 and 3 Unified
10 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Pipe and
Phase 1:
Phase 2:
Phase 3:

Preliminaries
Rules
Further filtering
Phase 3: Further filtering

Filter 0. (Optional) Places are only upper-case words.
Strong assumption (e.g. false for automatic transcriptions).
Filter 1. Places are not known FPs.

User-defined dictionary with found FPs.
Andare in tilt. (Rule 3)
Filter 2. Places are not (conjugated) verbs.

Stemming of candidate words. Filtered if recognised as a verb.
Il ministro esce camminando lentamente. (Rule 3)
Remaining positives are accepted as Locations and given to

the Ontology Model.
11 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
The Ontologies: Places Modelling
The Ontologies: Estimation Modelling
The Ontologies: Algorithms Modelling
Examples
Overview of the Semantic Web

The Semantic Web provides a common framework that allows
data to be shared and reused across application, enterprise, and
community boundaries Tim Berners-Lee
How data can be shared and reused?
Providing information an explicit meaning, so it can be
automatically processed and integrated by machines.
With automated reasoning procedures, implicit information in
data can extracted allowing to gain a deeper knowledge of the
domain.
Several tools are available.
12 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
What do we need?
A markup language such as XML.
IRI (Internationalized Resource Identifier).
Formal semantics and strict syntax.
Description language.
Description of the world we want to model (Ontology).
Reasoner and query language.
13 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
How we provide a description?

Thinking about the world in terms of Subject-Predicate-Object
(Triple).
Using default vocabularies to define set, individuals, datatypes,
and predicate features.
Combining our data with descriptions provided by other users.
Sharing data.
14 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples

What does Triple mean?
- Michelangelo made the Sistine Chapel ceiling
<http://wwww.unict.it/art.rdf#Michelangelo>
<http://wwww.unict.it/art.rdf#makes>
<http://wwww.unict.it/art.rdf#Sistine Chapel Ceiling>
- Michelangelo is an artist
<http://wwww.unict.it/art.rdf#Michelangelo>
<http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
<http://wwww.unict.it/art.rdf#Artist>
15 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
The Ontologies
OWL\SWRL.
Editor Protege.
Reuse LinkedGeoData ontology.
OpenStreetMap Dataset.
Compliant OWL\SWRL Reasoner such as Pellet.
OWL-API and JAVA to develop software interaction with
ontologies.
SPARQL query processor such as JENA.
16 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
17 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
Modelling Localisation Estimation

A candidate on the text.
An estimation (method and tools used).
One or more match with respective degree of belief.
Geographical places for every match.
Greatest degree, best result
What if estimation changes?
18 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
19 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
20 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
21 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
Examples
22 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
Examples
23 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
Examples
24 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography

The Ontologies
Examples
Examples
25 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Future Works
Extending the algorithm with other rules and filtering;

Extending both algorithm and ontology with Context
Sensitivity;
Using a local dataset including also data from local
government or final users;
Applying to Digital Humanities: OntoCeramic and Catania
Benedectine Monastery (San Nicol
o lArena).
26 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Bibliography
E. Agichtein and L. Gravano. Snowball: Extracting relations from large
plain-text collections. In Proceedings of the fifth ACM conference on Digital
libraries. ACM, 2000, pp. 8594.
A. Ballatore, D. C. Wilson, and M. Bertolotto. A Survey of Volunteered Open
Geo-Knowledge Bases in the Semantic Web. Quality Issues in the Management
of Web Information, ISRL 50, pp. 93120, Springer, 2013.
D. Cantone, C. Longo, M. Nicolosi-Asmundo, and D. F. Santamaria, Web
ontology representation and reasoning via fragments of set theory, in Web
Reasoning and Rule Systems - 9th International Conference, RR 2015, Berlin,
Germany, August 4-5, 2015, Proceedings, 2015, pp. 61-76.
D. Cantone, M. Nicolosi-Asmundo, D. F. Santamaria, and F. Trapani.
Ontoceramic: an OWL ontology for ceramics classification. In Proc. of the 30th
Italian Conference on Computational Logic, CILC 2015, Genova, Italy, July 1-3,
2015, CEUR Workshop Proceedings, ISSN 1613-0073, vol. 1459, pp. 122127.
D. Caruso, R. Giunta, D. Messina, G. Pappalardo, and E. Tramontana.
Rule-based location extraction from italian unstructured text. In Proc. of the
16th Workshop From Objects to Agents, WOA 2015, Naples, Italy, June
17-19, 2015, CEUR Workshop Proc., ISSN 1613-0073, vol. 1382, pp. 4652.
27 / 28
Introduction
ASM Algorithm
OWL Framework
Future Works
Bibliography
Bibliography
J. Lafferty, A. McCallum, and F. C. Pereira, Conditional random fields:
Probabilistic models for segmenting and labeling sequence data. In Proceedings
of the Eighteenth International Conference on Machine Learning (ICML), 2001,
pp. 282289.
LinkedGeoData, linkedgeodata.org/.
Lyndon J. B. Nixon, R. Volz, F. Ciravegna, and R. Studer. Ontology based
entity disambiguation with natural language patterns. Fourth International
Conference on Digital Information Management, ICDIM 2009, November 1-4,
2009, University of Michigan, Ann Arbor, Michigan, USA, pp. 1926.
K. Nigam, J. Lafferty, and A. McCallum. Using maximum entropy for text
classification. In IJCAI-99 workshop on machine learning for information
filtering, vol. 1, 1999, pp. 6167.
Ontology Web Language, http://www.w3.org/2001/sw/wiki/OWL.
OpenStreetMap, www.openstreetmap.org/.
S. Sarawagi. Information extraction. Foundations and trends in databases, vol.
1, no. 3, pp. 261377, 2008.
28 / 28

An OWL Framework For Rule-Based Recognition of Places in Italian Non-Structured Text

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

An OWL Framework For Rule-Based Recognition of Places in Italian Non-Structured Text

Hochgeladen von

Copyright:

Verfügbare Formate

Introduction

An OWL framework for rule-based recognition of

KDWEB 2016 - Cagliari, September 9, 2016

KDWEB 2016 - Cagliari, September 9, 2016

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

Semantic Web: Overview

Different approaches were studied in the literature:

Linked data and ontologies have been used to address the

KDWEB 2016 - Cagliari, September 9, 2016

Recognition of location names belonging to the Italian

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Pipe and Filter Multi-Agent Model

Three phases, connected in a pipe & filter style.

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Longer texts, e.g. a novel, are split into sentences.

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Phase 2: Rule 1 (Da Roma)

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Phase 2: Rule 2 & Rule 3

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Phase 2: Rule 1, 2 and 3 Unified

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Filter Multi-Agent Model

Phase 3: Further filtering

Filter 1. Places are not known FPs.

Filter 2. Places are not (conjugated) verbs.

Remaining positives are accepted as Locations and given to

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

Overview of the Semantic Web

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

Overview of the Semantic Web

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

Overview of the Semantic Web

How we provide a description?

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

Overview of the Semantic Web

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

Cantone, Fornaia, Nicolosi-Asmundo, Santamaria, Tramontana

KDWEB 2016 - Cagliari, September 9, 2016

Semantic Web: Overview

The Ontologies: Places Modelling