Automatic Acquisition of Hyponyms from Large Text Corpora

By Marti A. Hearst. 1992

We describe a method for the automatic acquisition of the hyponymy lexical relation from unrestricted text. Two goals motivate the approach: (i) avoidance of the need for pre-encoded knowledge and (ii) applicability across a wide range of text. We identify a set of lexico-syntactic patterns that are easily recognizable, that occur frequently and across text genre boundaries, and that indisputably indicate the lexical relation of interest. We describe a method for discovering these patterns and suggest that other lexical relations will also be acquirable in this way. A subset of the acquisition algorithm is implemented and the results are used to augment and critique the structure of a large hand-built thesaurus. Extensions and applications to areas such as information retrieval are suggested.

Subfields: 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 Natural Language Processing

Citations in the corpus

Semantic Parsing on Freebase from Question-Answer Pairs (2013, methodology): “We perform a similar procedure that uses a Hearst-like pattern (Hearst, 1992) to map phrases to unary predicates.”
Snorkel: Rapid Training Data Creation with Weak Supervision (2017, background): “For example, pattern-based heuristics encompass feature annotations [51] and pattern-bootstrapping approaches [18,20] (Example 2.”
Learning How to Ask: Querying LMs with Mixtures of Soft Prompts (2021, background): “In general, there may be many varied lexical patterns that signal a particular relation, and having more patterns will get better coverage (Hearst, 1992; Riloff and Jones, 1999).”
Social Bias Frames: Reasoning about Social and Power Implications of Language (2019, methodology): “We collect free-text answers in the form of simple Hearst-like patterns (e.g., “women are ADJ”, “gay men VBP”; Hearst, 1992).”
Distant supervision for relation extraction without labeled data (2009, methodology): “We use 1.2 million Wikipedia articles and 1.8 million instances of 102 relations connecting 940,000 entities.”
Ordinal Common-sense Inference (2016, background): “Many have pursued automating the process, such as in expanding lexical hierarchies (Hearst, 1992; Snow et al., 2006), constructing inference patterns (Lin and Pan-tel, 2001; Berant et al., 2011), reading reference materials (Richardson et al., 1998; Suchanek et al., 2007), mining search engine…”
Automatically Solving Number Word Problems by Semantic Parsing and Reasoning (2015, methodology): “To improve coverage, we run a term peer similarity and hypernym extraction al-gorithm (Hearst, 1992; Shi et al., 2010; Zhang et al., 2011) on a web snapshot of 3 billion pages, and get a peer-similarity graph and a collection of is-a pairs.”
Survey of the State of the Art in Natural Language Generation: Core tasks, applications and evaluation (2017, methodology): “This list of potential similes is then used to create Google queries, inspired by the work of Hearst (1992), of the form n-like t (e.g., stick insect-like Paris Hilton, which actually occurs on the web), giving a ranking of the potential similes to be generated.”
SemEval-2010 Task 8: Multi-Way Classification of Semantic Relations Between Pairs of Nominals (2009, methodology): “at the second one; other researchers (Kim and Baldwin, 2005; Nakov and Hearst, 2008; Nastase et al., 2006; Turney, 2005; Turney and Littman, 2005) have used their class scheme and data set. Moldovan et al. (2004) propose a 35-class scheme to classify”
Distributional Memory: A General Framework for Corpus-Based Semantics (2010, positive): “For instance, NOUN such as NOUN and such NOUN as NOUN were ﬁrst proposed by Hearst (1992) as highly reliable patterns for hypernym identiﬁ-cation, whereas (the) attribute noun of (a | the) NOUN is ADJ and (a | the) ADJ attribute noun of NOUN were successfully used to identify typical values of…”
Inherent Disagreements in Human Textual Inferences (2019, methodology): “Our co-hyponyms are taken froman internal database, which we constructed by running Hearst patterns (Hearst, 1992) over a large text corpus.”
Domain and Function: A Dual-Space Model of Semantic Relations and Compositions (2012, methodology): “Like past researchers (Rosario & Hearst, 2001; Rosario et al., 2002; Nastase & Szpakowicz, 2003; Veale, 2004; Nastase et al., 2006), we look for high values of simf(a, c) and simf(b, d) as indicators that simr(a :b, c :d) should be high, but we also look for high values of simd(a, b) and simd(c, d).”
QED: A Framework and Dataset for Explanations in Question Answering (2020, background): “…keeping with its modularity, we can extend QED to account for these by looking to semantic relations beyond referential equality and predicate entailment, such as set-membership noun phrase (Hearst, 1992) and interclausal (Miltsakaki et al., 2004; Lamm et al., 2018; Tandon et al., 2019) relations.”
Ontology-based information extraction: An introduction and a survey of current approaches (2010, background): “Such patterns include Hearst patterns like “<CONCEPT>s such as <INSTANCE>” [50].”
Do Supervised Distributional Methods Really Learn Lexical Inference Relations? (2015, background): “These features are relics of the Hearst (1992) pattern “y such as x”.”
Corpus-based Learning of Analogies and Semantic Relations (2005, methodology): “Thus the phrase “the bow lute, such as the Bambara ndang” suggests that the Bambara ndang is a type of (hyponym of) bow lute (Hearst, 1992).”
Learning Syntactic Patterns for Automatic Hypernym Discovery (2004, background): “The second classifier consists of only the " NP and/or other NP " subset of Hearst's patterns, as used in the automatic construction of a noun-labeled hypernym taxonomy in [1] .”