Robust Textual Inference via Graph Matching

By Aria Haghighi, Andrew Y. Ng, Christopher D. Manning. 2005

198 citations recorded by the source corpus.

We present a system for deciding whether a given sentence can be inferred from text. Each sentence is represented as a directed graph (extracted from a dependency parser) in which the nodes represent words or phrases, and the links represent syntactic and semantic relationships. We develop a learned graph matching approach to approximate entailment using the amount of the sentence's semantic content which is contained in the text. We present results on the Recognizing Textual Entailment dataset (Dagan et al., 2005), and show that our approach outperforms Bag-Of-Words and TF-IDF models. In addition, we explore common sources of errors in our approach and how to remedy them.

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

A Decomposable Attention Model for Natural Language Inference (2016, background): “A large body of work based on neural networks for text similarity tasks including NLI have been published in recent years (Hu et al., 2014; Rockt¨aschel et al., 2016; Wang and Jiang, 2015; Yin et al., 2016, inter alia ).”
Stress Test Evaluation for Natural Language Inference (2018, methodology): “The NLI task attracted significant interest before datasets became large enough for the application of neural methods (Glickman et al., 2005; Harabagiu and Hickl, 2006; Romano et al., 2006; Dagan et al., 2006; Giampiccolo et al., 2007; Dagan et al., 2010; MacCartney, 2009; Zanzotto et al., 2006; Mal”
Semantic Parsing via Paraphrasing (2014, background): “Many existing paraphrase models introduce latent variables to describe the derivation of c from x , e.g., with transformations (Heilman and Smith, 2010; Stern and Dagan, 2011) or alignments (Haghighi et al., 2005; Das and Smith, 2009; Chang et al., 2010).”
EQUATE: A Benchmark Evaluation Framework for Quantitative Reasoning in Natural Language Inference (2019, background): “…has attracted community-wide interest as a stringent test for natural language understanding (Cooper et al., 1996; Fyodorov; Glickman et al., 2005; Haghighi et al., 2005; Harabagiu and Hickl, 2006; Romano et al., 2006; Dagan et al., 2006; Giampiccolo et al., 2007; Zanzotto et al., 2006;…”
Learning to Grade Short Answer Questions using Semantic Similarity Measures and Dependency Graph Alignments (2011, methodology): “Haghighi et al. (2005), working on the entailment detection problem, point out that ﬁnding a good alignment is not sufﬁcient to determine that the aligned texts are in fact entailing.”
Learning to recognize features of valid textual entailments (2006, methodology): “Haghighi et al. (2005) divide the search into two steps: in the ﬁrst step they consider node scores only, which relaxes the problem to a weighted bipartite graph matching that can be solved in polynomial time, and in the second step they add the edges scores and hillclimb the alignment via an…”
Knowledge-Based Textual Inference via Parse-Tree Transformations (2015, background): “Systems based on lexical (word-based or phrase-based) matching of h in t (Haghighi et al., 2005; MacCartney, Galley, & Manning, 2008) or on heuristic transformation of t into h (Kouylekov & Magnini, 2005; Harmeling, 2009) typically apply only lexical rules (without variables), where both sides of…”
AdvEntuRe: Adversarial Training for Textual Entailment with Knowledge-Guided Examples (2018, unknown): “However, these also require model-specific engineering (Raina et al., 2005; Haghighi et al., 2005; Silva et al., 2018).”
Discriminative Learning over Constrained Latent Representations (2010, background): “There are several exceptions to the two-stage approach in the NLP community (Haghighi et al., 2005; McCallum et al., 2005; Goldwasser and Roth, 2008b; Das and Smith, 2009); however, the intermediate representation and the inference for constructing it are closely coupled with the application task.”
Dependency-based Semantic Role Labeling of PropBank (2008, methodology): “It has also been used in prototypes of NLP systems that carry out complex reasoning, such as entailment recognition systems (Haghighi et al., 2005; Hickl et al., 2006).”
Posterior Regularization for Structured Latent Variable Models (2010, background): “They are useful for language modeling, textual entailment and machine translation (Haghighi et al., 2005; Chelba et al., 1997; Quirk et al., 2005; Shen et al., 2008), to name a few tasks.”
A Survey of Answer Extraction Techniques in Factoid Question Answering (2006, background): “(Haghighi, Ng, and Manning 2005) proposed an approach by representing each sentence in the pair as a directed acyclic graph, where the nodes are lexical words and edges are dependency relations between words.”
A Survey of Paraphrasing and Textual Entailment Methods (2009, background): “Paraphrases may also be recognized by computing similarity measures on graphs whose edges do not correspond to syntactic dependencies, but reflect semantic relations mentioned in the input expressions (Haghighi, 2005), for example the relation between a buyer and the entity bought.”
A Discourse Commitment-Based Framework for Recognizing Textual Entailment (2007, background): “…et al., 2006), (2) measures of term overlap (Jijkoun and de Rijke, 2005), (3) the alignment of graphs created from syntactic or semantic dependencies (Haghighi et al., 2005), or (4) statistical classi-ﬁers which leverage a wide range of features, including the output of paraphrase generation…”
Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance? (2020, background): “NLI is a benchmark task for research on natural language understanding (Cooper et al., 1996; Haghighi et al., 2005; Harabagiu and Hickl, 2006; Dagan et al., 2006; Giampiccolo et al., 2007; Zanzotto et al., 2006; MacCartney, 2009; Dagan et al., 2010; Marelli et al., 2014).”
Syntactic/Semantic Structures for Textual Entailment Recognition (2010, methodology): “Shallower models exploit matching methods between syntactic/semantic graphs of texts and hypotheses (Haghighi et al., 2005).”