Semantic Roles and Relations

1
Semantic Roles and Relations

• Nikhil Pandey
• 04D05011

2
Introduction

• Universal grammar
• Natural Language
• UNL
• UNL Relations
• Universal words UWs
• Semantic roles
• Examples

3
UNL

• It is an interlingua that has been proposed by
  the United Nations University to access, transfer
  and process information on the internet in the
  natural languages of the world. UNL represents
  information sentence by sentence.

4
UNL Overview

• UWs
• Relations
• Example
• I can hear a man crying outside.
• unl                                             
             
• aoj(hear(iclgtperceive(agtgtthing,objgtthing))._at_entry
  ._at_ability,  I)
• obj(hear(iclgtperceive(agtgtthing,objgtthing))._at_entry
  ._at_ability,  01)  
• agt01(cry(agtgtman)._at_entry,  man(iclgtmammal))
                           
• plc01(cry(agtgtman)._at_entry,  outside(iclgtplace))
                 
• /unl

5
Universal Words

• Headword
• Constraints
• Attribute list
• Example

6
UNL Relations

• Conditions
• Necessary
• Sufficient
• Example of the relation agt
• A person
• A machine
• A thing

7
Restrictions

• UWs have three restrictions
• Knowledge base restrictions
• Semantic restrictions
• Argument frame restrictions

8
Semantic categories of UW

• Verbal Concept
• Adjectival Concept

9
Semantic Roles

• System for identifying the semantic
  relationships, or semantic roles, filled by
  constituents of a sentence within a semantic
  frame.
• Current information extraction and dialogue
  understanding systems are based on domain
  specific frame-and-slot templates.

10
Frames

• Each frame contains roles
• Judgement(Judge, Evaluee, Reason)
• Statement(Speaker, Addressee, Message)
• and not always do the frames contain all their
  roles.
• Eg. Holman would characterize this as blaming
  the poor
• the poor is the Evaluee.

11
Semantic Roles types

• Domain-specific roles (such as from-airport,
  to-airport, depart-time, landing -time).
• Or Verb-specific roles (Eater and Eaten for
  eat).
• Or Proto-roles and Macro-roles
• Or Fillmore (Agent, Experiencer, Instrument,
  Object, Source, Goal, Location, Time, Path)
• Since there is not always a direct correspondence
  between Semantics and Syntax, it is hard to
  automatically identify a Frame element.

12
FrameNets

• FramNet project proposes roles that are neither
  general nor specific. The FrameNet roles are
  defined for each specific frame.
• A frame is a schematic representation of
  situations involving various participants, props
  and other conceptual role.
• Example The frame Conversation is invoked by
  related verbs (argue, debate, converse) or
  nouns(dispute or discussion)

13
Features used in assigning Semantic roles

• Phrase Type
• Governing Category
• Parse Tree Path
• Position
• Voice
• Head Word

14
Example

• S
• NP
  VP
• PRP VBD NP
  SBAR
  
• 
  IN
  S
• 
  
  NP VP
• 
  NNP
  VBD NP PP
• 
  
  PRP IN NP
• 
  
  
  
• 
  
  
  NN
  
• He heard the sound of liquid slurping in a
  metal container as Farrell
  approached him from behind

15
Example Parse Tree Path

• S
• NP
  
  VP
• PRP
  VB
  NP
• DT NN
• He
  ate
  some pancakes
• VB VP S NP is the path from ate to he

16
Quantifier Phrase

• S
• QP
  SBAR
• PDT IN NP
  VP
• PRP
  NNPS VBD JJ
• Both of his
  legs got injured

17
Empty category

• 
  S
• NP
  VP
• 
  VBD NP S
• 
  PRP NP
  VP
• 
  
  TO VP
• 
  
  VB NP
• 
  
  DT NN
• 
  1
• I
  forced him to
  give a party

18
Penn Treebank part-of-speech tags

• CC Coordin. Conjunction (and, but, or)
• CD Cardinal Number (numbers)
• DT Determiner WDT Wh- Determiner (Which,
  that)
• EX Existensial there (there)
• FW Foreign Word
• IN Preposition
• JJ adjective JJR comparative
  adjective JJS Superlative Adjective
• LS- List item marker (1, 2, one)
• MD- Modal (can, should)
• NN Noun NNS plural NNP Proper and
  singular NNPS- Proper and prular
• PRP Personal pronoun PRP - Possessive
  pronoun(your, ones) WP Wh- Pronouns(what,
  who)
• RB- Adverb RBR Comparative RBS
  Superlative
• UH Interjection
• VB - verb VBD past tense verb VBG
  gerund(eating) VBN past participle

19
Constituent/Non- terminal Labels

• ADJP Adjective Phrase
• ADVP Adverb Phrase
• CONJP Conjunction Phrase
• SBAR Clause introduced by a complementizer
• SBARQ Question introduced by WH- word
• NP Noun Phrase
• VP Verb Phrase
• PP Prepositional Phrase
• FRAG Fragment
• INTJ - Interjection
• QP Quantifier Phrase

20
CONCLUSION

• CONCLUSION

21
REFERENCES

• 1 Universal Networking Language (UNL)
  Specifications. UNL Center of UNDL Foundation.
• URL http//www.undl.org/unlsys/unl/unl2005/.
•  
• 2 Automatic Labeling of Semantic Roles. Gildea
  Daniel and Jurafsky Daniel.
•  
• 3 Chapters 2 and 3 from the UNL Reference
  Manual, Igor Boguslavsky
•  
• 4 Wikipedia URLs
• http//en.wikipedia.org/wiki/Universal_Networking_
  Language
• http//en.wikipedia.org/wiki/Natural_Language
• http//en.wikipedia.org/wiki/Semantic_Roles

22
THANK YOU

• Q A