A General Framework for Distributional Similarity

By Julie Weeds, David J. Weir. 2003

155 citations recorded by the source corpus.

We present a general framework for distributional similarity based on the concepts of precision and recall. Different parameter settings within this framework approximate different existing similarity measures as well as many more which have, until now, been unexplored. We show that optimal parameter settings outperform two existing state-of-the-art similarity measures on two evaluation tasks for high and low frequency nouns.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning

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 Representation & Self-Supervised Learning

Citations in the corpus

From image descriptions to visual denotations: New similarity metrics for semantic inference over event descriptions (2014, background): “More recently, asymmetric similarities have been proposed as more suitable for semantic inference tasks such as entailment (Weeds and Weir, 2003; Szpektor and Dagan, 2008; Clarke, 2009; Kotlerman et al., 2010).”
Learning Entailment Rules for Unary Templates (2008, methodology): “Weeds and Weir (2003) proposed to measure the symmetric similarity between two words by averaging two directional (asymmetric) scores: the coverage of each word’s features by the other.”
Focused Entailment Graphs for Open IE Propositions (2014, methodology): “One of the more indicative similarity measures, inclusion (Weeds and Weir, 2003), yielded only 27% precision at 40% recall when tuning a threshold to optimize F 1 .”
The Distributional Inclusion Hypotheses and Lexical Entailment (2005, background): “Each feature represents another word (or term) with which w co-occurs, and possibly specifies also the syntactic relation between the two words as in (Grefenstette, 1994; Lin, 1998; Weeds and Weir, 2003).”
Directional distributional similarity for lexical inference (2010, methodology): “Another such common symmetric measure is the widely cited and competitive (as shown in Weeds and Weir 2003) LIN measure (Lin 1998a), deﬁned as where F x is the feature vector of a term x , w x ( f ) is the weight of the feature f in that term’s vector, set to their pointwise mutual information…”
Crowdsourcing Inference-Rule Evaluation (2012, methodology): “Lastly, we computed the area under the recall-precision curve (AUC) for DIRT , Cover , BInc and Berant et al. ’s method, resulting in an AUC of 0.4, 0.43, 0.44, and 0.52 respectively.”
Characterising Measures of Lexical Distributional Similarity (2004, methodology): “The ﬁnal three measures are settings in the additive MI-based Co-occurrence Retrieval Model (AMCRM) (Weeds and Weir, 2003; Weeds, 2003).”
Do Supervised Distributional Methods Really Learn Lexical Inference Relations? (2015, background): “Many unsupervised distributional methods of recognizing lexical inference replace cosine similarity with an asymmetric similarity function (Weeds and Weir, 2003; Clarke, 2009; Kotlerman et al., 2010; Santus et al., 2014).”
On the Systematicity of Probing Contextualized Word Representations: The Case of Hypernymy in BERT (2020, background): “Additionally, our work ties into the rich history on modeling hypernymy in NLP systems (Lin, 1998; Weeds and Weir, 2003; Baroni et al., 2012; Rimell, 2014; Roller et al., 2014; Weeds et al., 2014; Shwartz et al., 2015; Vuli´c and Mrkˇsi´c, 2018) and evaluating distributional semantic models on…”
Context Mover's Distance & Barycenters: Optimal transport of contexts for building representations (2018, background): “The early unsupervised approaches for this task exploited different linguistic properties of hypernymy (Weeds & Weir, 2003; Kotlerman et al., 2010; Santus et al., 2014; Rimell, 2014).”