Discovery of inference rules for question-answering

By Dekang Lin, Patrick Pantel. 2001

640 citations recorded by the source corpus.

One of the main challenges in question-answering is the potential mismatch between the expressions in questions and the expressions in texts. While humans appear to use inference rules such as ‘X writes Y’ implies ‘X is the author of Y’ in answering questions, such rules are generally unavailable to question-answering systems due to the inherent difficulty in constructing them. In this paper, we present an unsupervised algorithm for discovering inference rules from text. Our algorithm is based on an extended version of Harris’ Distributional Hypothesis, which states that words that occurred in the same contexts tend to be similar. Instead of using this hypothesis on words, we apply it to paths in the dependency trees of a parsed corpus. Essentially, if two paths tend to link the same set of words, we hypothesize that their meanings are similar. We use examples to show that our system discovers many inference rules easily missed by humans.

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

Annotation Artifacts in Natural Language Inference Data (2018, methodology): “, 2012) is severely biased towards distributional methods, since it was created using DIRT (Lin and Pantel, 2001).”
Constructing Datasets for Multi-hop Reading Comprehension Across Documents (2017, background): “…interaction between question, context, answer candidates and combinations thereof (Peng et al., 2015; Weissenborn, 2016; Xiong et al., 2017; Liu and Perez, 2017), which is often iterated over several times (Sordoni et al., 2016; Neumann et al., 2016; Seo et al., 2017b; Hu et al., 2017) and may…”
Distant supervision for relation extraction without labeled data (2009, methodology): “…or unsupervised IE has used only simple lexical features, thereby avoiding the computational expense of parsing (Brin, 1998; Agichtein and Gravano, 2000; Etzioni et al., 2005), and the few systems that have used unsupervised IE have not compared the performance of these two types of feature.”
Paraphrasing Revisited with Neural Machine Translation (2017, background): “Examples include query and pattern expansion (Riezler et al., 2007), summarization (Barzilay, 2003), question answering (Lin and Pantel, 2001), semantic parsing (Berant and Liang, 2014), semantic role labeling (Woodsend and La-pata, 2014), and machine translation (Callison-Burch et al., 2006).”
The Stanford Typed Dependencies Representation (2008, background): “Prepositions often work as role markers, and this type of link facilitates the extraction of how the two content words are related; and thus these links are often used by downstream applications (Lin and Pantel, 2001; Snow et al., 2005).”
ParaNMT-50M: Pushing the Limits of Paraphrastic Sentence Embeddings with Millions of Machine Translations (2017, methodology): “Many methods have been developed for generating or finding paraphrases, including using multiple translations of the same source material (Barzilay and McKeown, 2001), using distributional similarity to find similar dependency paths (Lin and Pantel, 2001), using comparable articles from multiple new”
Bootstrapping Path-Based Pronoun Resolution (2006, methodology): “Our method for determining path coreference is similar to an algorithm for discovering para-phrases in text (Lin and Pantel, 2001).”
Transforming Question Answering Datasets Into Natural Language Inference Datasets (2018, methodology): “This is the same type signature as for sentence simpliﬁca-tion (Chandrasekar et al., 1996), paraphrase (Lin and Pantel, 2001; Bannard and Callison-Burch, 2005) and summarization (Jones, 1993), high-lighting the close connection between these tasks.”
A Probabilistic Classification Approach for Lexical Textual Entailment (2005, background): “Recent works on paraphrase and entailment rule acquisition (Barzilay and McKeown 2001; Lin and Pantel 2001; Szpektor et al. 2004) learn lexical syntactic paraphrase patterns.”
A study of the knowledge base requirements for passing an elementary science test (2013, methodology): “, NELL’s earlier use of N-FOIL to induce Horn clauses [12], or DIRT’s use of distributional similarity [13].”
Corpus-based and Knowledge-based Measures of Text Semantic Similarity (2006, methodology): “A related line of work consists of methods for paraphrase recognition, which typically seek to align sentences in comparable corpora (Barzilay & Elhadad 2003; Dolan, Quirk, & Brockett 2004), or paraphrase generation using distributional similarity applied on paths of dependency trees ( Lin & Pantel ”
Semantic Matching in Search (2014, methodology): “Lin and Pantel [125] propose a method of extracting synonymous patterns such as “X is author of Y” is equivalent to “X wrote Y”, by parsing a large number of texts and mining similar paths of parsed trees in”
Knowledge-Based Textual Inference via Parse-Tree Transformations (2015, methodology): “…(Barzilay & McKeown , 2001; Barzilay & Lee, 2003; Pang, Knight, & Marcu, 2003), by querying the Web (Ravichandran & Hovy, 2002; Szpektor et al., 2004), or from a local corpus (Lin & Pantel, 2001; Glickman & Dagan, 2003; Bhagat & Ravichandran, 2008; Szpektor & Dagan, 2008; Yates & Etzioni, 2009).”
Semantic Inference at the Lexical-Syntactic Level for Textual Entailment Recognition (2007, methodology): “In order to find lexical-syntactic paraphrases and entailment rules, such as ‘X strike Y → X hit Y ’ and ‘X buy Y →X own Y ’ that would bridge between p and h, we applied the DIRT algorithm (Lin and Pantel, 2001) to the first CD of the Reuters RCV1 corpus1.”
Investigating a Generic Paraphrase-Based Approach for Relation Extraction (2006, methodology): “Several methods extract atomic paraphrases by exhaustively processing local corpora (Lin and Pantel, 2001; Shinyama et al., 2002).”
Learning Entailment Rules for Unary Templates (2008, methodology): “This sparked intensive research on unsupervised acquisition of entailment rules (and similarly paraphrases) e.g. (Lin and Pantel, 2001; Szpektor et al., 2004; Sekine, 2005).”
The Excitement Open Platform for Textual Inferences (2014, methodology): “Particularly, the EOP platform includes a language independent tool to build Wikipedia resources (Shnarch et al., 2009), as well as a language-independent framework for building distributional similarity resources like DIRT (Lin and Pantel, 2002) and Lin similarity(Lin, 1998).”