Svetla Koeva, Stoyan Mihov

1
Semantic Relations in INTEX

• Svetla Koeva, Stoyan Mihov
• 6th INTEX Workshop, Sofia, Bulgaria
• 30 May, 2003

2
The main idea

• We describe a method for presenting the synonymy
  and hyperonymy relations in Natural Languages
  with the INTEX system.
• Creation of specialized semantic dictionaries for
  different semantic relations. Those dictionaries
  are extracted automatically using
• the WordNet structure and
• inflectional (DELAF/DELACF type) dictionaries.

3
WordNet

• WordNet is an electronic lexical thesaurus based
  on word meanings rather than word forms.
• By means of semantic relations, all word meanings
  in a language can be interconnected, constituting
  a huge network - WordNet.
• WordNet has been initially developed at Princeton
  for English.
• The success of WordNet has determined the
  emergence of several projects that aim at the
  development of WordNets for languages other than
  English.
• EuroWordNet, BakaNet

4
The notion of synset

• A synset is a set of words with the same
  part-of-speech that can be interchanged in a
  certain context.
• Synonymy is symmetric, transitive relation of
  equivalence.
• The synsets are equivalence classes of the
  synonymy relation.
• For example the set car, auto, automobile,
  machine, motorcar forms a synset.

5
Hyperonymy / Hyponymy and other semantic relations

• Hyperonymy and hyponymy are inverse, asymmetric
  and transitive relations. The relation implies
  that the hyperonym may substitute the hyponym in
  a context but not the other way around.
• A ????? (dish) is a kind of ????? (food)
• A ????? (food) is a kind of ????? (dish)
• Synsets can be related to each other by many
  other semantic relations, such as meronymy
  (between parts and wholes), anthonymy, etc.

6
The semantic relations in INTEX

• Semantic relations such as symonymy, hyperonymy,
  etc. can be presented and exploited with the
  INTEX system.
• For each semantic relation we create a specific
  semantic dictionary using the DELAF/DELACF
  framework.

7
Structure of the semantic dictionaries

• The dictionary consists of pairs of words in the
  corresponding semantic relation.
• car,automobile.N
• auto,automibile.N
• Additionally we have to express in the dictionary
  all corresponding word forms their synonymy.
• car,automobile.N
• cars,automobile.N
• auto,automibile.N
• autos,automibile.N
• Those dictionaries are extracted automatically
  using the WordNet structure and inflectional
  dictionaries. The same technique is applied for
  both DELAF and DELAFC dictionaries.

8
Available semantic dictionaries

• Synonymy dictionaries for the first order base
  concepts containing about 1300 synsets in
  Bulgarian and English
• Bulgarian DELAF synonymy dictionary 26231
  inflected entries
• English DELAF synonymy dictionary 77830
  inflected entries
• Sample hyperonymy dictionaries for English and
  Bulgarian

9
Applications of the Semantic Dictionaries

• Verification of consistency and completeness of
  the WordNet Databases
• Information retrieval by means of semantic
  equivalence with synonymy dictionaries
• Information retrieval by means of semantic
  specification with hyperonymy dictionaries.

10
Application of Synonymy dictionary for Bulgarian
Locate pattern lt???gt
11
Application of Synonymy dictionary for English
Locate pattern ltcasegt
12
Application of Hyperonymy dictionary
Locate pattern ltvehiclegt
13
Problems

• Semantic ambiguity different meanings of the
  same word can lead to wrong extractions.
• Processing of parallel texts in two languages.
• The transitive closure of the transitive closure
  can connect too many words to a given concept.

14
Future directions

• Extensions and enhancements of the semantic
  dictionaries by means of
• Extension of the dictionaries coverage
• Addition of other semantic relations
• Inclusion of additional information to the
  entries.
• Development of semantic disambiguation rules.
• Integration of multilingual semantic extraction
  with INTEX using the Inter-Lingual-Index
  relation.