1ISWC, 2009

2

Ontologies (conceptualizations of the world): No world ontology: multiple domain ontologiesHard and costly to developharder and more costly to adapt to unintended uses

(semantic heterogeneity problem)Even harder and more costly to maintain in timeSuccessful in niche areasMostly simplified ontologies (e.g., lightweight

ontologies)(very ) limited success of OWLSKOS used only in limited ways in digital librariesDid not scale to becoming the ultimate solution to

the knowledge anda data management problem

3

Tags (annotations of objects, eg. photos, docs, persons, ..): Used a lotBillions of tags Largely used in the WebFrom tags to metadata (data about data)Large scale (meta)data bases exist (DBPEDIA, YAGO, …)RDF is the language for representing metadataRDF is scaling to the WebLimited success of RDF as implementation languageRDF supported by most vendorsProblems of qualityProblems of semantic heterogeneityFrom (meta)data to open linked (government) (meta)data

4

Entities (objects with a name, e.g., persons, locations, events, music, books, monuments, …): From a data centric to an entity centric vision of the world (and the Web)Entities as information (meta-data) aggregators Use and extension of existing standards (DublinCore, FOAF, eventML, …)An entity centric vision of the Web from information retrieval to information extractionThe Internet of things (entities are also in the world)from open linked data to open linked entities

http://entitypedia.org University of Trento

Search and navigation on entities Semantic or faceted search on entities From search to question answering Summarization of entities and concepts Entity matching Automatic population of a target ontology with

entities CRUD on entities and their metadada Towards the goal of becoming a hub towards

indexing (parts of) the Web

http://entitypedia.org University of Trento

Entity Types: different kinds of entities and

entity-specific attributes Identity Attributes: the attributes that allow to

uniquely identify an entity w.r.t. the other ones Faceted search and systematic way of organizing

facets in domains Multilinguality: starting from Italian and English

languages Controlled user customizable Dictionary A never reached before accuracy and quality of

the data (type checking)

http://entitypedia.org University of Trento

Entitypedia: some figures

SOURCE CONCEPTS RELATIONS ENTITIES (*)

WordNet 2.1 110,609 204,481 -

MultiWordNet 33,356 - -

GeoNames 663 ~ 7 Millions ~ 7 Millions

YAGO 2009 - ~ 1.1 Millions

Planned to grow thanks to the Wisdom of crowds

http://entitypedia.org soon to come!!!!

http://entitypedia.org University of Trento

BUT

Role of space (Semantic heterogeneity problem) Role of time (evolution of data and metadata) … and not only

The central problem of managing diversity

11ISWC, 2009

12

Plague or Delicacy?Thanks to Claudia!

13

14

In languageHow many names do you have for snow? (the role of weather)“Bug as disease” vs.“bug as food” (the role of domains)

In data “Transportation is on foot” vs.“transportation is by plane” (the role of time)“The President is Obama” vs. “the President is Berlusconi” (the role of space)

In knowledge“There are 2 types of music: traditional and modern” vs. “there are 50 types of music further refined in 100 types (pop, pop-country, …) (the role of goals/ needs/ competence)

In opinions“Bugs are great food” vs.“how can you eat bugs?” (the role of culture)“Climate is/ is not an important issue” (the role of schools of thought)

15

The main cause of the semantic gap between our globalized conceptualizations of the world, expressed using language, and our local experience of the world, whose most direct representations are media,

Diversity as a bug (up to the early Web). The current implementation of the web is an “implementational mistake”: we can pretend it is like querying a centrally designed data base

Diversity as a must (the Semantic Web). Diversity is unavoidable, it is the reason for diverging viewpoints and conflicts: we need semantics in order to “absorbe” diversity and reduce it to the centrally designed data base approach

Diversity as a feature. Diversity is a local maximum: we should make it traceable, understandable, and use it to develop better technology, e.g., diversity aware classification, navigation and search in large scale, long living (eternal), heterogeneous multimedia datasets (e.g., the Web of to day)

The LIVING WEB

16

The world, our experience about the world, our data and knowledge about the world are strongly influenced by diversity in, e.g., geographical contexts, weather and time of the day, cultural backgrounds, schools of thought, ... and many others.

Time and evolution add a further dimension making diversity an even further intrinsic and unavoidable property of the world, and our data and knowledge of the world.

Diversity is a local optimum!

We envisage a future where data and knowledge management

tools (implementing, e.g., search, navigation, reasoning, ...) will trace, understand, exploit diversity in very large multimedia datasets (in particular, the Web itself) and, therefore, will produce more insightful, better organized, easier-to-understand output.

17

September 15, 2008 Living Knowledge Negotiation meeting 18

Property (P)

Infection

DiseasePreliminariesObstetrics

General

Poison

VirusBacteria

Venereal disease

Tuberculosis

Functional disorder

Bacterial

Special modifier (m)

Disease modifier

Infectious

Viral Fungal

Space modifier (m)

India

Asia

World

Europe

Italy France

Entity (E)

Knee

Tissue

Body and its organs Circulatory

systemDigestive system

Cell

Lower extremity

Foot Leg

Toe

Head

Pathology

Action (A)

Nursing

Symptom and diagnosis

Clinical

Physical method

Microscope

X-ray

Cell viruses

Diseases

Clinical diagnosis in Italy

Localization = context dependent application ofthe mechano property on global domains and facets

Property (P)

dangerous

wild

Space modifier (m)

India

Asia

World

Europe

Italy France

Entity (E)

Knee

feline

tiger

Foot Leg

Toe

Action (A)

walking

A wild tiger walking in India

21

Ex.: <Id, Drills, Cutting machine (other), subsumes>

… The big picture

September 15, 2008 Living Knowledge Negotiation meeting 22

Individual Universal

community

Individual Universal community

25

F. Giunchiglia, “Contextual reasoning”, Epistemologia - Special Issue on I Linguaggi e le Macchine, 1993.

C. Ghidini, F. Giunchiglia: Local models semantics, or contextual reasoning = locality + compatibility. Artificial Intelligence Journal, 127(3), 2001.

F. Giunchiglia, I.Zaihrayeu: Making peer databases interact – a vision for an architecture supporting data coordination. CIA’02

P. Bernstein, F. Giunchiglia, A. Kementsietsidis, J. Mylopoulos, L. Serafini, and I. Zaihrayeu: Data management for Peer-to-Peer Computing, WebDB’02

Bouquet, F. Giunchiglia, F. van Harmelen, L. Serafini, H. Stuckenschmidt: C-OWL: Contextualizing Ontologies, ISWC 2003.

M. Bonifacio, F. Giunchiglia, I. Zaihrayeu: Peer-to-Peer Knowledge Management, I-KNOW 2005

F. Giunchiglia, “Managing Diversity in Knowledge”, Invited talk, Proceedings ECAI 2006. Presentation available on line.

Giulia Boato, Claudio Fontanari, Fausto Giunchiglia, Francesco De Natale, Glocal Search, DISI Technical Report, 2008

September 15, 2008 Living Knowledge Negotiation meeting 26