Formal analysis of concepts and onomasiological search of terms: towards new dictionaries

1
Formal analysis of concepts and onomasiological
search of terms towards new dictionaries

• Amparo Alcina
• TecnoLeTTr_at_
• Universitat Jaume I de Castellón

TAMA Computational Terminology Workshop 9-10
October 2008
2
Index

• Introduction
• Translators and terminologists needs
• Onomasiological dictionaries
• ONTODIC
• Aims
• Context
• Techniques
• Applications

3
Introduction
4
Justification

• Increase of specialized translation
• Globalization
• More specialized subjects
• More terminological units and new concepts
• Terminological needs for human and machine
  translation

Specialized texts
terminology
Translated texts
5
Translators needs

• Translators use complex strategies that combine
• Textual corpora and concordances
• Texts in the Internet
• Queries in definitions and other fields in
  electronic dictionaries
• Comparison of definitions in monolingual
  dictionaries

6
Problem

• Lack of terminological dictionaries
• Current dictionaries offer semasiological
  inquiries
• Fireman -gt Person who works candling fires.
• But do not offer onomasiological inquiries
• A lot of trees in a mountain -gt forest

7
manada cerdo piara
Herd pork herd2
8
By subject (science and technic)
9
Reverse dictionary, i.e. fire
http//dictionary.reference.com/reverse/
10
Problems for elaborating terminologies

• Terminological databases -gt Limitations as
• How to handle the branches of the subject tree
• How to handle different kinds of concepts
  properties, relations, groups of terms
• How to build terminological definitions
• Production of definitions by hand -gt not
  efficient
• Production of definitions by copy and paste -gt
  low quality

11
Three reasons for ONTODIC

• Necessity of onomasiological search for
  translation purposes
• Terminological databases do not offer concept
  management facilities
• There is a lack of computer assisted terminology
  tools to support concept analysis

12
Aims and Techniques in ONTODIC
13
Context

• Terminology Theory and terminography
• Restricted domain
• Only one or few senses for a term
• Computational Linguistics and Knowledge
  Engineering
• Ceramics terminology
• Knowledge, documentation, experts

14
Phases

• Conceptual description of terms
• Formalize the conceptual analysis
• Implement the conceptual analysis

15
Terminological description of concept terms

• Each term must be assigned to a branch in the
  domain tree
• Terms in a branch must be organized in concept
  groups
• Each concept group must to share the same
  description template
• Each description template contains
• Concept class (descriptor)
• Features of the term
• Properties
• Relationships

16
Ceramics domain tree

• 1. Producto
• 1.1 Producto acabado
• Características del prod. Acabado
• Defectos del prod. Acabado
• Tipos de producto acabado
• Partes del producto acabado
• 1.2.Producto intermedio
• Características del prod. intermedio
• Defectos del prod. intermedio
• Tipos de producto
• 1.3. Otros productos
• 2. Fabricación
• Procesos
• Maquinaria y accesorios
• 3. Materias primas y aditivos
• Materias primas
• Aditivos

• 4. Colocación
• 4.1. Procesos de colocación
• 4.2. Maquinaria y accesorios de colocación
• 4.3. Materiales para colocación
• 4.3.1. Materiales de base
• 4.3.2. Materiales de adherencia
• 5. Calidad
• 5.1. Procesos de control de calidad
• 5.2. Maquinaria y accesorios de control de
  calidad
• 6. Unidades de medida
• 7. Organismos, Instituciones

17
Concept groups Manufacture
production
Prensado explosivo
distribution
paletización
empaquetamiento
Prensado isostático
preparation
amasado
atomización
18
Analysis of terminological concepts

• Term
• Subject
• Branch
• Concept group
• Descriptor
• Characteristics

19
Exemple of terminological concept description in
normal database

• Term tamizado
• Subject cerámica
• Branch Fabricación
• Concept group proceso de preparación
• Descriptor proceso de preparación
• Characteristics
• el material pasa a través de un tamiz que
  retiene los elementos más gruesos

20
Relationships between terms
21
Clasify elements of description

• el material pasa a través de un tamiz que
  retiene los elementos más gruesos
• The material go through a sieve that retains
  the large elements
• Patient the material
• Action go through a sieve
• Result retains the large elements

22
Generalize (not ad hoc)

• Estabilishing a broad, high level classification
  of terms, for instance, a meta-ontology (Sager
  and Kageura 1995)
• Establishing a broad classification of
  relationships in terminology with formal
  specifications (Maroto 2007)
• Establishing a broad classification of properties
  of terms (Estellés 2006 based on Kageura)

23
Concept classes

• Entities
• Activities
• Relations
• Properties

24
Concept classes

• Entities
• products, machines, materials
• Activities
• production, preparation
• Relations
• part-whole, generic-species
• Properties
• product characteristics

25
Relation template
26
Exemple of relation description
27
Exemple of search

• a (domain) R b (rango)
• Which kinds of ceramic tiles are produced by
  prensado en seco?
• a (ACTIVITY) R (process-product) b (ENTITY)
• (Prensado en seco) process-product (ceramic tile)

28
Characteristic template
29
Exemple of Characteristic description
30
Implementing

• Concept analysis tools
• Terminology
• CODE, COGNITERM (Meyer 1997)
• OntoTerm (Faber)
• CAOS
• Knowledge Engineering
• Ontology editors WebODE, OntoEdit, KAON, Protégé

• Protégé
• University of Standford
• Free
• Open
• A lot of functions, plug-ins
• XML, OWL
• inquiries

31
Advantages of Protégé

• Similarity to terminological elements
• Domain tree
• Hierarchical organization of terms
• Relations
• Characteristics
• Inquiries

32
Problems

• Many elements must be adapted to our needs
• The structure should be useful not only for a
  subject area, but also for other subjects
• We must also work with denominations, for
  instance, equivalences

33
Exemple of search

• a (domain) R b (rango)
• Which kinds of ceramic tiles (b) are produced by
  prensado en seco (a)?
• a (ACTIVITY) R (process-product) b (ENTITY)
• (Prensado en seco) produce (ceramic tile)

34
Example of implementation
35
Exemple of query
36
Example of implementation
37
Applications

• Offer tools for lexicographers, terminologists
  and translators, developing the concept analysis
  in an easily manageable way
• Offer tools and terminological resources for
  queries ussually made by translators
• Offer resources for Natural Language Processing
  (NLP)

38
Muchas gracias por vuestra atención!

• Amparo Alcina (alcina_at_trad.uji.es)
• http//tecnolettra.uji.es
• Universitat Jaume I
• Spain