Test Suites for Textual Inference - PowerPoint PPT Presentation

1 / 9
About This Presentation
Title:

Test Suites for Textual Inference

Description:

Test Suites for Textual Inference. Johan Bos. University of Rome ... Nominal Anaphora. Ellipsis. Adjectives. Comparatives. Temporal Reference. Verbs. Attitudes ... – PowerPoint PPT presentation

Number of Views:49
Avg rating:3.0/5.0
Slides: 10
Provided by: joha171
Category:

less

Transcript and Presenter's Notes

Title: Test Suites for Textual Inference


1
Test Suites for Textual Inference
  • Johan BosUniversity of Rome La Sapienza
  • Websitehttp//www.cogsci.ed.ac.uk/jbos/ti/

2
Textual Inference Data Sets
  • Pascal data sets
  • RTE 1 (1.376 pairs)
  • RTE 2 (1.600 pairs)
  • Other less known data sets
  • FRACAS (346 pairs) Cooper et al. 1996
  • PARC (76 pairs) Zaenen, Karttunen Crouch 2005
  • Specific phenomena
  • Adjectives (ca. 1.000 pairs) Amoia Gardent
    2006

3
The FRACAS test suite
  • European Project on Computational Semantics, in
    the mid 1990s
  • Test suite published in D16, but since then
    forgotten?
  • Cooper et al. (1996) Using the Framework. Fracas
    deliverable D16, section 3
  • Aim of test suite measure semantic competence of
    NLP system

4
The FRACAS test suite
  • Grouped on linguistic and semantic phenomena
  • Generalised quantifiers
  • Plurals
  • Nominal Anaphora
  • Ellipsis
  • Adjectives
  • Comparatives
  • Temporal Reference
  • Verbs
  • Attitudes

5
FRACAS example pairs
  • 3.209 Mickey is a small animal.
    Dumbo is a large animal.
    Is Mickey
    smaller than Dumbo? YES
  • 3.205 Dumbo is a large animal.
    .. ... Is Dumbo
    a small animal? NO
  • 3.206 Fido is not a small animal.
    .. Is Fido a
    large animal? DONT KNOW

6
Differences
  • Textual inference, not textual entailment
  • Instead of a two valued classification TRUE and
    FALSE, we have three values YES, NO, DONT KNOW
  • No or very little world knowledge
  • Focussing on specific phenomena
  • Not real text, but artificially constructed

7
Discussion
  • Advantages
  • Useful for debugging NLP systems
  • Better control over whether a system understands
    certain phenomena
  • Similar proposal
  • Zaenen, Karttunen Crouch 2005
  • Opponents of this proposal
  • Manning 2006

8
The way forward
  • Not one single task, but
  • Introduce different classes of RTI
  • Linguistic/Semantic
  • World knowledge
  • Robust
  • Limit artificiality of examples
  • Construct T from real text
  • Construct H from template rules

9
Example
  • Text, from real data
  • The Osaka World Trade Center is the highest
    building in Japan
  • Generated Hypotheses
  • The Osaka World Trade Center is the third highest
    building in Japan NO
  • The Osaka World Trade Center is a building in
    Japan YES
  • The Osaka World Trade Center is one of the
    highest buildings in Japan YES
  • The Osaka World Trade Center is the highest
    building in Western Japan MAYBE
Write a Comment
User Comments (0)
About PowerShow.com