Contradictions and Justifications: Extensions to the Textual Entailment Task
Voorhees, Ellen M.

Article Structure

Abstract

The third PASCAL Recognizing Textual Entailment Challenge (RTE-3) contained an optional task that extended the main entailment task by requiring a system to make three-way entailment decisions (entails, contradicts, neither) and to justify its response.

Introduction

The PASCAL Recognizing Textual Entailment (RTE) workshop series (see www.pascal—network .

The Three-way Decision Task

The extended task used the RTE-3 main task test set of entailment pairs as its test set.

J ustifications

The second part of the extended task was for systems to provide explanations of how they reached their conclusions.

Conclusion

The RTE-3 extended task provided an opportunity to examine systems’ abilities to detect contradiction and to provide explanations of their reasoning

Topics

NIST

Appears in 17 sentences as: NIST (18)
In Contradictions and Justifications: Extensions to the Textual Entailment Task
  1. The answer key for the three-way decision task was developed at the National Institute of Standards and Technology ( NIST ) using annotators who had experience as TREC and DUC assessors.
    Page 2, “The Three-way Decision Task”
  2. NIST assessors annotated all 800 entailment pairs in the test set, with each pair independently annotated by two different assessors.
    Page 2, “The Three-way Decision Task”
  3. The three-way answer key was formed by keeping exactly the same set of YES answers as in the two-way key (regardless of the NIST annotations) and having NIST staff adjudicate assessor differences on the remainder.
    Page 2, “The Three-way Decision Task”
  4. Main Task NIST Judge 1
    Page 3, “The Three-way Decision Task”
  5. Main Task NIST Judge 2 YES UN KN N 0
    Page 3, “The Three-way Decision Task”
  6. Recall that NIST judges annotated all 800 entailment pairs in the test set, with each pair independently annotated twice.
    Page 3, “The Three-way Decision Task”
  7. For each entailment pair, one of the NIST judges was arbitrarily assigned as the fi rst judge for that pair and the other as the second judge.
    Page 3, “The Three-way Decision Task”
  8. The agreement between NIST and RTE annotators is shown in Table 2.
    Page 3, “The Three-way Decision Task”
  9. The NIST judges’ answers are given in the columns and the two-way reference answers in the rows.
    Page 4, “The Three-way Decision Task”
  10. Agreement is then computed as the percentage of matches when a NIST judge’s NO or UNKNOWN annotation matched a NO two-way reference answer.
    Page 4, “The Three-way Decision Task”
  11. The 90% agreement rate between the NIST judges and the two-way answer key probably reflects a somewhat larger amount of disagreement since the test set already had RTE annotators’ disagreements removed.
    Page 4, “The Three-way Decision Task”

See all papers in Proc. ACL 2008 that mention NIST.

See all papers in Proc. ACL that mention NIST.

Back to top.