A Phrase-Based Alignment Model for Natural Language Inference

By Bill MacCartney, Michel Galley, Christopher D. Manning. 2008

139 citations recorded by the source corpus.

The alignment problem---establishing links between corresponding phrases in two related sentences---is as important in natural language inference (NLI) as it is in machine translation (MT). But the tools and techniques of MT alignment do not readily transfer to NLI, where one cannot assume semantic equivalence, and for which large volumes of bitext are lacking. We present a new NLI aligner, the MANLI system, designed to address these challenges. It uses a phrase-based alignment representation, exploits external lexical resources, and capitalizes on a new set of supervised training data. We compare the performance of MANLI to existing NLI and MT aligners on an NLI alignment task over the well-known Recognizing Textual Entailment data. We show that MANLI significantly outperforms existing aligners, achieving gains of 6.2% in F1 over a representative NLI aligner and 10.5% over GIZA++.

Subfields: Artificial Intelligence, 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 Natural Language Processing

Citations in the corpus

A Decomposable Attention Model for Natural Language Inference (2016, background): “, 2013), natural language inference (Marsi and Krahmer, 2005; MacCartney et al., 2006; Hickl and Bensley, 2007; MacCartney et al., 2008), and semantic parsing (Andreas et al.”
Hypothesis Only Baselines in Natural Language Inference (2018, methodology): “In turn, NLI models often used alignment-based techniques (MacCartney et al., 2008)”
Learning Natural Language Inference with LSTM (2015, background): “, 2009), perform explicit sentence alignment (MacCartney et al., 2008) or use formal logic (Clark and Harrison, 2009).”
An extended model of natural logic (2009, methodology): “We have investigated this problem separately in (MacCartney et al. 2008).”
Learning Answer-Entailing Structures for Machine Comprehension (2015, methodology): “This is the structure considered in a majority of previous works in RTE (MacCartney et al., 2008) and QA (Yih et al., 2013) as they only reason on single sentence length texts.”
Diversify Your Datasets: Analyzing Generalization via Controlled Variance in Adversarial Datasets (2019, methodology): “While earlier models for these tasks relied on domain knowledge and lexical resources like WordNet (e.g MacCartney et al., 2008; Heilman and Smith, 2010), the release of the large-scale Stanford natural language inference dataset (SNLI; Bowman et al., 2015) shifted the focus to neural models which…”
Discriminative Learning over Constrained Latent Representations (2010, methodology): “Examples include RTE (Zanzotto and Moschitti, 2006; MacCartney et al., 2008; Roth et al., 2009); string matching (Bergsma and Kon-drak, 2007); transliteration (Klementiev and Roth, 2008); and paraphrase identiﬁcation (Qiu et al., 2006; Wan et al., 2006).”
Natural Logic and Natural Language Inference (2014, background): “We have investigated this problem separately in [12].”
A Survey of Paraphrasing and Textual Entailment Methods (2009, methodology): “30See chapter 4 of the book “Speech and Language Processing” (Jurafsky & Martin, 2008) and chapter 6 of the book “Foundations of Statistical Natural Language Processing” (Manning & Schuetze, 1999) for an introduction to language models. SRILM (Stolcke, 2002) is a commonly used tool to create languag”
Data-Driven Response Generation in Social Media (2011, background): “Alignment is more difﬁcult in conversational data than bilingual data (Brown et al., 1990), or textual entailment data (Brockett, 2006; MacCartney et al., 2008).”
MonaLog: a Lightweight System for Natural Language Inference Based on Monotonicity (2019, methodology): “Unlike existing work on natural logic, however, our model avoids the need for having expensive alignment and search sub-procedures (MacCartney et al., 2008; Stern and Dagan, 2011), and relies on a much smaller set of background knowledge and primitive relations than MacCartney and Manning (2009).”
Word Alignment by Fine-tuning Embeddings on Parallel Corpora (2021, methodology): “Previous methods (MacCartney et al., 2008; Yao et al., 2013a,b; Sultan et al., 2014) typically exploit external resources such as WordNet to tackle the problem.”
Lexicosyntactic Inference in Neural Models (2018, background): “…be-tween clause-embedding verbs, negation, and embedded clause type – is important for any system that aims to do general natural language inference (MacCartney et al. 2008 et seq ; cf. Dagan et al. 2006) or event extraction (see Grishman and Sund-heim 1996 et seq ), and it seems unlikely to be…”
Neural Natural Language Inference Models Enhanced with External Knowledge (2017, background): “Unlike early work on NLI (Jijkoun and de Rijke, 2005; MacCartney et al., 2008; MacCartney, 2009) that explores external knowledge in conventional NLI models on relatively small NLI datasets, we aim to merge the advantage of powerful modeling ability of neural networks with extra external inference…”