Adriana Roventini*

1
Adriana Roventini Rita Marinelli
Extending the Italian WordNet with the
Specialized Language of the Maritime
Domain Istituto di Linguistica Computazionale
del CNR Pisa Italye-mail rita.marinelli_at_ilc.
cnr.it - adriana.roventini_at_ilc.cnr.it
2
Our purpose
to describe the construction we are carrying
out at the Institute for Computational
Linguistics, of a terminological subset belonging
to the maritime lexical domain (in particular to
the technical and commercial/maritime transport
domain).
3
Wordnet

• In the Princeton semantic WordNet (Miller et al.,
  1990) the meanings of words are represented in
  terms of their conceptual-semantic and lexical
  relations to other words
• it has been the tool of choice for building
  Natural Language Processing (NLP) systems of
  various kinds.

4
EWN

• The main goals of the EuroWordNet (EWN) are
• to develop a (multilingual) lexical resource,
  retaining the basic underlying design of WordNet
  1.5 (hereafter WN1.5)
• to improve it in order to meet the needs of
  research in the field of NLP (Vossen, 1999).

5
Background

• SI-TAL an Italian national Project (Integrated
  System for the Automatic Treatment of Language)
• development of various integrated language
  resources and software tools for the automatic
  treatment of Italian written and spoken language
• ITALWORDNET lexical semantic resource developed
  within the SI-TAL project, enlarging the first
  database built in EWN.

6
IWN

• EWN project IWN
  SI-TAL
• 
  Integrated System for the Automatic
• 
  
  Treatment of Language

• IWN database containing ca. 50.000 synsets
• Nouns
• Verbs
• Adjectives
• Adverbs
• Proper Names
• IWN links synsets by lexical-semantic relations
• Synonymy
• the most important
  relations
• Hyponymy
• Many other semantic relations encoded for various
  subsets of Italian Nouns (Common Proper ),
  Verbs, Adjectives
• IWN synsets linked toWordNet 1.5 through a
  generic ILI (InterLingual
• Index)

Not encoded in EWN
7
The IWN linguistic model

• Synsets and synonymy relation
• Synset as basic notion
• around which WN, EWN and IWN are built
  synset or set of synonymous words belonging to
  the same Part-of-Speech (PoS) that can be
  interchanged at least in a context.
• Synsets are connected by semantic relations
  to other synsets and to the ILI (an unstructured
  version of WN 1.5, containing all its synsets but
  not the relations among them).

8
Inherited from EWN also

• language-internal relations link the
  language-specific synsets (mainly
  hyperonymy/hyponymy or is-A relation, role,
  causes, purpose, part relations, etc.)
• equivalence relations link the Italian synsets
  to the InterLingual-Index (ILI).
• By linking our wordnet to the ILI we ensured the
  possibility to use IWN for multilingual
  applications.

9
Reasons for our choice

• The globalisation of trade, business and travel
  and the technology development (growing
  importance of transport).
• The changes produced within the maritime activity
  and the related terminology (remarkable incidence
  of this lexical domain)
• New techniques of communication, translation and
  diffusion of terms (monopole of the English
  language).

10
Building/structuring the terminological IWN

• according to the design principles of the generic
  wordnet, (applying the same semantic relations
  model)
• exploiting the possibility - available in IWN
  through the Inter-Lingual Index (ILI) - of
  linking the specialized terms to the
  corresponding closest concepts in English.

11
Sources

• Several information sources have been used to
  select the BC
• the Dizionario Globale dei termini marinareschi,
  edited by the Capitaneria del Porto di Livorno,
  online on the Web
• the Dizionario di marina, edited by Barberi
  Squarotti G. , Gallinaro I, (2002)
• the Glossario dello spedizioniere (Annuario
  Federspedi 1988)
• the Dizionario di termini marittimi mercatili,
  compiled by P. R. Brodie and translated by E.
  Vincenzini, Lloyds of London Press, Legal
  Publishing and Conferences Division, 1988.

12
Choice of the base concepts (BCs)

• design of the terminological database top level,
  identifying the most relevant and representative
  domain concepts or basic concepts (BCs) .
• (i.e. showing a large number of hyponyms,
  and/or more frequently used in this particular
  domain of maritime navigation and transport).

13
First Base-Concepts

• A first nucleus of over 200 BCs was identified,
  such as nave (ship), porto (harbour), ormeggio
  (mooring), albero (mast), carico (cargo),
  spedizione (shipment), navigazione (navigation),
  trasporto (transport), tariffa (tariff), nolo
  (freight) and so on, which are sufficiently
  general and constitute the root nodes of the
  specialized database.

14
BCs export/import

• as XML files
• (see the example below concerning the verb
  imbarcare/to ship).
•  

IWN
xml
IWNTerm
15

• Example of an XML export file imbarcare (to
  ship)
• - ltWORD_MEANING ID"V32560" PART_OF_SPEECH"V"gt
•   ltGLOSS /gt
• - ltVARIANTSgt
•   ltLITERAL LEMMA"imbarcare" SENSE"1"
  STATUS"CT" /gt
•   lt/VARIANTSgt
• - ltINTERNAL_LINKSgt
• - ltRELATION TYPE"xpos_near_synonym" ID"2"
  INV_ID"2"gt
•   ltTARGET_WM ID"27869" PART_OF_SPEECH"N"
  LEMMA"imbarco" SENSE"1" GLOSS"" /gt
•   lt/RELATIONgt
• - ltRELATION TYPE"has_hyperonym" ID"8"
  INV_ID"8"gt
•   ltTARGET_WM ID"32127" PART_OF_SPEECH"V"
  LEMMA"fare" SENSE"14" GLOSS"causare un
  cambiamento in un processo o uno stato (seguito
  da un infinito)." /gt
•   lt/RELATIONgt
• - ltRELATION TYPE"has_hyponym" ID"10"
  INV_ID"10"gt
•   ltTARGET_WM ID"36489" PART_OF_SPEECH"V"
  LEMMA"reimbarcare" SENSE"1" GLOSS"" /gt
•   lt/RELATIONgt
• - ltRELATION TYPE"involved_instrument" ID"31"
  INV_ID"31"gt
•   ltTARGET_WM ID"15111" PART_OF_SPEECH"N"
  LEMMA"imbarcatoio" SENSE"1" GLOSS"" /gt
•   lt/RELATIONgt

16
New BCs

• Other BCs were included ex novo, not present
  with their maritime senses in the generic
  database, but very frequently used and
  representative of this specific domain, for
  instance nolo (freight), classe (class), fanale
  (light), punto (position), destino (destination),
  agente marittimo (shipping agent), spedizioniere
  (freight forwarder).

17
Example Punto (Position)
18
Use of Relations to codify specialized terms

• first nucleus of terms increased
• (encoding hyponyms and using other semantic
  relations)

19
Example Ormeggio (Mooring)
20
Kind and Number of Terms

• 2227 lemmas corresponding to 1721 synsets and
  2355 word-senses belonging to the maritime
  (technical/nautical and maritime transports)
  domain all linked to the generic wordnet.
• Terms belonging to all the different grammatical
  categories of nouns, verbs, adjectives, adverbs
  and a small set of proper names have been
  codified in the terminological data base (3971
  relations).

21
Example Porto (Harbour)
22
Polilexical Units

• Base Concepts (BCs) as the root of a
  terminological sub-hierarchy
• (in many cases) hyponyms BC adjective or
  prepositional phrase
• For instance
• carico (cargo),
• carico completo (full cargo), carico di
  merci varie (general cargo), carico in coperta
  (deck cargo), carico parziale (part load cargo),
• tariffa (tariff),
• tariffa doganale (custom tariff), tariffa di
  trasporto (transport tariff), tariffa forfettaria
  (flat-rate tariff),
• nolo (freight)
• nolo anticipato (freight prepaid), nolo
  intero (full freight), nolo secondo il valore (ad
  valorem freight), nolo a destino (freight payable
  at destination).

23
Linking Terms to the ILI

• Actually the English term or multiword (or its
  acronym) is often known and used much more than
  the Italian one in the maritime transport
  activity.
• Difficulty in finding the synonyms
• both the English term (or multiword) and the
  Italian
• one are included in the synset as variants,
  (as we thought
• this could be useful to non-professionals as
  well).

24
EXAMPLES

• RO-RO (Roll On/Roll Off) usually indicates nave
  traghetto per automezzi (ferry for vehicles
  transport),
• the abbreviation FOB (Free On Board) is used to
  say con le spese pagate fino a bordo, (loading
  costs paid up to the ships broadside),
• CIF (Cost Insurance and Freight) to say costi
  fino a bordo più assicurazione e nolo mare pagati
  (loading costs, insurance and sea-freight
  prepaid).

25
The Link Structure

• the BCs identified for this terminological
  lexicon constitute the top level and are the root
  nodes for the plug-in operation which allows
  linking between the generic and the specialized
  wordnet.

26
Two types of plug_in relations are codified

• the eq-plug-in relation, as equivalence
  synonymy relation between synsets of the two
  databases
• the has-hyperonym(hyponym)-plug relation, as
  equivalence hyperonymy/hyponymy relation between
  synsets of the two databases.

27
Tool Facilities

• a simultaneous parallel consultation of the two
  databases to facilitate insertion of the
  relations
• an integrated research between the two databases
  
• if the lemma is found in both databases and
  there is an eq-plug-in relation between the
  synsets, the synset belonging to the specific
  domain eclipses the generic one exploiting the
  integrated research.

28
Tool Facilities
downward and horizontal relations (part-of
relations, role relations, cause relations,
derivation, etc.) are taken from the
terminological wordnet. upward (hyperonymy)
relations are taken from the generic one. It is
possible to access the generic database or the
terminological database or both databases at the
same time.
29
EXAMPLE Nolo (Freight)
30
Nolo plug-in (with downward relations)
31
Nolo plug-in (with upward relations)
32
EXAMPLE Bussola (Compass)
33
Bussola plug_in (with downward relations)
34
Bussola plug_in (with upward relations)
35
Differences between IWN and Dictionaries/Glossarie
s

• The data are not only described (by the
  definition), but also codified (by relations)
• data structured only alfabetically in the
  dictionary edited by the Harbour Master (we can
  read for example all information about bussola
  all together and almost confused) become, in a
  relational database, synsets, linked to each
  other by many types of semantic relations
  (hyperonymy, hyponymy, holo/mero part, etc.)
  which can also be managed automatically.

36
FINAL REMARKS

• maritime terminology is object of great interest
  in a maritime nation like Italy, which has a
  strong marine tradition
• the English terms prevail over the Italian
  synonyms
• maritime terminology dictionaries are rare and
  sometimes it is very difficult to find an English
  translation of these terms

37
Instrument for work

• The possibility of having definitions and
  translations of specific terms is a useful
  instrument for work (export-import companies,
  maritime agencies, etc.), at school and the
  didactic activities of various types (nautical
  Institutes, professional training, etc.) and, in
  general, whenever a reference to terms of this
  specific domain is needed.

38

• From a commercial point of view, the English
  language prevails over all other languages
  contracts, negotiations, chartering and operation
  documents of cargo ships (like bills of lading,
  etc.) are in English, and so are a great number
  of reference books.
• from the point of view of usefulness, there are
  circumstances in which it is necessary to refer
  to a translation of technical terms that is
  correct, abreast and absolutely unambiguous.

39
Our aim

• to build a terminological database showing the
  semantic relations between different concepts, a
  precise correct linkage to the English terms, and
  then to make it a point of reference, in
  circumstances like legal actions, for instance,
  when the judge..
• to carry on this research increasing the number
  of terms and starting a cooperation with the
  official transport organizations in order to
  enrich and refine this product and to arrive at a
  definitive version recognized and validated.
• to start this kind of research for the Italian
  language.

40
Results

• Specialized lexicon enlarged
• Italian terms clarified
• More effective management of Italian terms and
  English terms
• In spite of globalisation, in a maritime
  country like ours it is absolutely essential not
  to lose our linguistic identity