Difference between revisions of "Textual Entailment Resource Pool"

Revision as of 08:11, 12 October 2012

Textual entailment systems rely on many different types of NLP resources, including term banks, paraphrase lists, parsers, named-entity recognizers, etc. With so many resources being continuously released and improved, it can be difficult to know which particular resource to use when developing a system.

In response, the Recognizing Textual Entailment (RTE) shared task community initiated a new activity for building this Textual Entailment Resource Pool. RTE participants and any other member of the NLP community are encouraged to contribute to the pool.

In an effort to determine the relative impact of the resources, RTE participants are strongly encouraged to report, whenever possible, the contribution to the overall performance of each utilized resource. Formal qualitative and quantitative results should be included in a separate section of the system report as well as posted on the talk pages of this Textual Entailment Resource Pool.

Adding a new resource is very easy. See how to use existing templates to do this in Help:Using Templates.

Complete RTE Systems

VENSES (from Ca' Foscari University of Venice, Italy)
Nutcracker (available for download)
Entailment Demo (from the University of Illinois at Urbana-Champaign) - INACTIVE (as of 2010-12-22)
EDITS - Edit Distance Textual Entailment Suite (open source software developed by Human Language Technology (HLT) group at FBK-Irst)
BIUTEE - Bar Ilan University Textual Entailment Engine (open source).

RTE data sets

Past campaigns data sets

RTE1 dataset - provided by PASCAL
RTE2 dataset - provided by PASCAL
RTE3 dataset - provided by PASCAL
RTE4 dataset - provided by NIST - freely available upon request. For details see TAC User Agreements
RTE5 dataset - provided by NIST - freely available upon request. For details see TAC User Agreements
RTE6 dataset - provided by NIST - freely available upon request. For details see TAC User Agreements
RTE7 dataset - provided by NIST - freely available upon request. For details see TAC User Agreements

Other data sets

FrameNet manually annotated RTE 2006 Test Set. Provided by SALSA project, Saarland University.
Manually Word Aligned RTE 2006 Data Sets. Provided by the Natural Language Processing Group, Microsoft Research.
RTE data sets annotated for a 3-way decision: entails, contradicts, unknown. Provided by Stanford NLP Group.
BPI RTE data set - 250 pairs, focusing on world knowledge. Provided jointly by Boeing, Princeton, and ISI.
Textual Entailment Specialized Data Sets - 90 RTE-5 Test Set pairs annotated with linguistic phenomena + 203 monothematic pairs (i.e. pairs where only one linguistic phenomenon is relevant to the entailment relation) created from the 90 annotated pairs. Provided jointly by FBK-Irst, and CELCT.
RTE-5 Search Pilot Data Set annotated with anaphora and coreference information - RTE-5 Search Data Set annotated with anaphora/coreference information + Augmented RTE-5 Search Data Set, where all the referring expressions which need to be resolved in the entailing sentences are substituted by explicit expressions on the basis of the anaphora/coreference annotation. Provided by CELCT and distributed by NIST at the Past TAC Data web page (2009 Search Pilot, annotated test/dev data).
RTE-3-Expanded, RTE-4-Expanded, RTE-5-Expanded. RTE data set expanded in the two and three way task, at least 2000 pairs in each data set.
Explanation-Based Analysis annotation of RTE 5 Main Task subset described in this ACL 2010 paper
Wiki Entailment Corpus A RTE-like set of entailment pairs extracted from Wikipedia revisions described in this paper
The Guardian Headlines Entailment Training Dataset An automatically generated dataset of 32,000 pairs similar to the RTE-1 dataset.

Knowledge Resources

The RTE Knowledge Resources page presents:

a call for resources, inviting system developers to share the resources used by their own TE engines, to both help improve the TE technology and further test and evaluate such resources;
the ablation tests carried out in the RTE challenges in order to evaluate the impact of knowledge resources and tools on TE system performances;
lists of knowledge resources, both publically available and unpublished, used by systems participating in the last RTE challenges.

Tools

Parsers

C&C parser for Combinatory Categorial Grammar
Minipar
Shallow Parser - from the University of Illinois at Urbana-Champaign, see a web demo of this tool

Role Labelling

ASSERT
Shalmaneser
Semantic Role Labeler - from the University of Illinois at Urbana-Champaign, see a web demo of this tool

Entity Recognition Tools

Illinois Named Entity Tagger - see a web demo of this tool
Illinois Multi-lingual Named Entity Discovery Tool - see a web demo of this tool

Similarity / Relatedness Tools

UKB: Open source WordNet-based similarity/relatedness tool, includes also pre-computed semantic vectors for all words

Corpus Readers

NLTK provides a corpus reader for the data from RTE Challenges 1, 2, and 3 - see the Corpus Readers Guide for more information.

Related Libraries

PyPES general purpose library containing evaluation environment for RTE and McPIET text inference engine based on the ERG (English Resource Grammar)

Text Normalizers

Java number normalizer (Beta) A tool for converting textual representations of numbers to a standard numerical string.

References

Links

@@ Line 25: / Line 25: @@
 * [http://www.nist.gov/tac/data/past/2008/RTE-4.html RTE4 dataset] - provided by [http://www.nist.gov/index.html NIST] - freely available upon request. For details see [http://www.nist.gov/tac/data/forms/index.html TAC User Agreements]
 * [http://www.nist.gov/tac/data/past/2009/RTE-5.html RTE5 dataset] - provided by [http://www.nist.gov/index.html NIST] - freely available upon request. For details see [http://www.nist.gov/tac/data/forms/index.html TAC User Agreements]
-* [http://www.nist.gov/tac/2010/RTE/index.html RTE6 dataset] - provided by [http://www.nist.gov/index.html NIST] - freely available upon request. For details see [http://www.nist.gov/tac/data/forms/index.html TAC User Agreements]
+* [http://www.nist.gov/tac/data/past/2010/RTE-6_Main_Task.html RTE6 dataset] - provided by [http://www.nist.gov/index.html NIST] - freely available upon request. For details see [http://www.nist.gov/tac/data/forms/index.html TAC User Agreements]
+* [http://www.nist.gov/tac/2011/RTE/index.html RTE7 dataset] - provided by [http://www.nist.gov/index.html NIST] - freely available upon request. For details see [http://www.nist.gov/tac/data/forms/index.html TAC User Agreements]
 === Other data sets ===
@@ Line 37: / Line 39: @@
 * [https://agora.cs.illinois.edu/display/rtedata/Explanation+Based+Analysis+of+RTE+Data Explanation-Based Analysis annotation of RTE 5 Main Task subset] described in [http://l2r.cs.uiuc.edu/~danr/Papers/SammonsVyRo10.pdf  this ACL 2010 paper]
 * [http://art.uniroma2.it/zanzotto/resources/WIKI_FINAL_CORPUS_v1.zip Wiki Entailment Corpus] A RTE-like set of entailment pairs extracted from Wikipedia revisions described in [http://aclweb.org/anthology/W/W10/W10-3504.pdf  this paper]
+* [https://github.com/daoudclarke/rte-experiment The Guardian Headlines Entailment Training Dataset] An automatically generated dataset of 32,000 pairs similar to the RTE-1 dataset.
 == Knowledge Resources ==