From Knowledge Organization (KO) to Knowledge Representation (KR) Fausto Giunchiglia, Biswanath Dutta, Vincenzo Maltese maltese@disi.unitn.it

Knowledge Organization: make information available 2 ISKO UK 11/22/2018 LIS developed its own very successful solutions for the classification and search of documents. In KO, search is focused on document properties (e.g. title, author, subject). KO tends to fail in situations when users express their needs in terms of entity properties (e.g., of the author) Knowledge Organization: make information available KO has developed its own solution for the classification and search of documents in libraries and digital archives

Limitations in KO: Lack of expressiveness 3 ISKO UK 11/22/2018 Search by document properties Give me documents with subject “Michelangelo” and “marble sculpture” Search by entity properties Give me documents about marble sculptures made by a person born in Italy Limitations in KO: Lack of expressiveness However, users may think by entity properties, instead of document properties.

Limitations in KO: Lack of formalization 4 ISKO UK 11/22/2018 Subject: Buonarroti, Michelangelo Subject: sculpture – Renaissance In indexes the syntax of the subjects is given by a subject indexing language grammar. What about the semantics? Is Michelangelo the Italian artist? When and where he was born? What are his most famous works? Is sculpture the form of art? Is Renaissance the historical period? When and where exactly? Limitations in KO: Lack of formalization Subjects are informal natural language strings, or more precisely their syntax is often formal (because of the subject indexing language grammar), but their semantics (the meaning) is left implicit.

From documents to entities 5 ISKO UK 11/22/2018 Name: David Class: statue Author: Michelangelo Date of creation: 1504 Matter: marble Height: 4.10 m From documents to entities Name: Michelangelo Surname: Buonarroti Class: artist Date of birth: 6th March 1475 Date of death: 18th February 1564 Thinking in terms of entities, we can describe works and people in terms of entities. This is done in LIS, but informally. One immediate consequence is that it seems there are not systematic ways to reason about these objects. As a matter of fact, information is scattered in many resources.

From KO to KR 6 ISKO UK 11/22/2018 KR search by any entity property KO search by title, author, subject In fact, documents are just on particular case of entities, with title, author and subject very important properties.

Classification Ontologies 7 ISKO UK 11/22/2018 Classification Ontologies

Descriptive Ontologies: intensional knowledge 8 ISKO UK 11/22/2018 Descriptive Ontologies: intensional knowledge

Descriptive Ontologies: extensional knowledge 9 ISKO UK 11/22/2018 Descriptive Ontologies: extensional knowledge

Descriptive Ontologies: extensional knowledge 10 ISKO UK 11/22/2018 Descriptive Ontologies: extensional knowledge Give me documents about any lake with depth greater than 100 written by Italians

11 ISKO UK 11/22/2018 How to build high quality and scalable descriptive ontologies? DERA is faceted as it is inspired to the principles and canons of the faceted approach by Ranganathan DERA is a KR approach as it models entities of a domain (D) by their entity classes (E), relations (R) and attributes (A) A new approach: DERA

Descriptive Ontologies in DERA (I) 12 ISKO UK 11/22/2018 Step 1: Identification of the atomic concepts   (E) watercourse, stream: a natural body of running water flowing on or under the earth Step 2: Analysis a body of water a flowing body of water no fixed boundary confined within a bed and stream banks larger than a brook Descriptive Ontologies in DERA (I) (1) We identify atomic concepts, synonyms and natural language definitions. Each concept is marked as E, R or A. (2) For each concept we identify its characteristics

Descriptive Ontologies in DERA (II) 13 ISKO UK 11/22/2018 Step 3: Synthesis. (E) Body of water (is-a) Flowing body of water (is-a) Stream (is-a) Brook (is-a) River (is-a) Still body of water (is-a) Pond (is-a) Lake   Step 4: Standardization. (E) stream, watercourse: a natural body of running water flowing on or under the earth Descriptive Ontologies in DERA (II) (3) Facets are built as descriptive ontologies (e.g., we make use of explicit is-a and part-of relations). (4) The preferred term is elected

Descriptive Ontologies in DERA (III) 14 ISKO UK 11/22/2018 Step 5: Ordering Terms and concepts in the facets are ordered   Step 6: Formalization Descriptive ontologies are translated into Description Logic formal ontologies, e.g.,: River ⊑ Stream River (Volga) Length (Volga, 3692) Descriptive Ontologies in DERA (III)

15 ISKO UK 11/22/2018 DERA facets have explicit semantics and are modeled as descriptive ontologies DERA facets inherits all the nice properties of the faceted approach, such as robustness and scalability DERA allows for a very expressive document search by any entity property DERA allows for automated reasoning via the formalization into Description Logics ontologies Properties of DERA

Conclusions The usefulness of moving from KO to KR 16 ISKO UK 11/22/2018 The usefulness of moving from KO to KR KO is methodologically very strong, but subjects are limited in formality and expressiveness as, by employing classification ontologies, it only supports queries by document properties. KR, by employing descriptive ontologies, supports queries by any entity property, but it is methodologically weaker than KO. We propose the DERA faceted KR approach DERA, being faceted, allows the development of high quality and scalable descriptive ontologies DERA, being a KR approach, allows modeling relevant entities of the domain and their E/R/A properties and enables automated reasoning. It supports a highly expressive search of documents exploiting entity properties. Conclusions

17 ISKO UK 11/22/2018 From KO to KR Thank you!