Learning to recognize features of valid textual entailments

By Bill MacCartney, Trond Grenager, Marie-Catherine de Marneffe, Daniel Matthew Cer, Christopher D. Manning. 2006

153 citations recorded by the source corpus.

This paper advocates a new architecture for textual inference in which finding a good alignment is separated from evaluating entailment. Current approaches to semantic inference in question answering and textual entailment have approximated the entailment problem as that of computing the best alignment of the hypothesis to the text, using a locally decomposable matching score. We argue that there are significant weaknesses in this approach, including flawed assumptions of monotonicity and locality. Instead we propose a pipelined approach where alignment is followed by a classification step, in which we extract features representing high-level characteristics of the entailment problem, and pass the resulting feature vector to a statistical classifier trained on development data. We report results on data from the 2005 Pascal RTE Challenge which surpass previously reported results for alignment-based systems.

Subfields: Artificial Intelligence, Molecular Biology, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

In Natural Language Processing

Citations in the corpus

An extended model of natural logic (2009, background): “To date, the most successful NLI systems have relied on surface representations and approximate measures of lexical and syntactic similarity to ascertain whether p subsumes h (Glickman et al. 2005, MacCartney et al. 2006, Hickl et al. 2006).”
Modeling Semantic Containment and Exclusion in Natural Language Inference (2008, methodology): “Our implementation of natural logic, the NatLog system, uses a multi-stage architecture like those of (Marsi and Krahmer, 2005; MacCartney et al., 2006), comprising (1) linguistic analysis, (2) alignment, (3) lexical entailment classification, (4) en-”
What is the Jeopardy Model? A Quasi-Synchronous Grammar for QA (2007, background): “We can also observe a trend in recent work in textual entailment that more emphasis is put on explicit learning of the syntactic graph mapping between the entailed and entailed-by sentences (MacCartney et al., 2006).”
Finding Contradictions in Text (2008, methodology): “Our system is based on the stage architecture of the Stanford RTE system (MacCartney et al., 2006), but adds a stage for event coreference decision.”
Natural Logic for Textual Inference (2007, methodology): “Our natural logic system, dubbed the NatLog system, has a three-stage architecture similar to those in (Marsi and Krahmer, 2005; MacCartney et al., 2006), comprising (1) linguistic pre-preprocessing, (2) alignment, and (3) entailment classification.”
A Phrase-Based Alignment Model for Natural Language Inference (2008, background): “If a high-scoring alignment indicates a close correspondence between H and P , does this also indicate a valid inference? We have previously emphasized (MacCartney et al., 2006) that there is more to inferential validity than close lexical or structural correspondence: negations, modals, non-factive”
Design and realization of a modular architecture for textual entailment (2013, background): “Often different aspects of quality are combined to arrive at a final decision using supervised learning methods (Monz and de Rijke 2001; MacCartney et al. 2006; Zanzotto, Pennacchiotti and Moschitti 2009).”