Finding Predominant Word Senses in Untagged Text

By Diana McCarthy, Rob Koeling, Julie Weeds, John A. Carroll. Annual Meeting of the Association for Computational Linguistics, 2004

373 citations recorded by the source corpus.

In word sense disambiguation (WSD), the heuristic of choosing the most common sense is extremely powerful because the distribution of the senses of a word is often skewed. The problem with using the predominant, or first sense heuristic, aside from the fact that it does not take surrounding context into account, is that it assumes some quantity of hand-tagged data. Whilst there are a few hand-tagged corpora available for some languages, one would expect the frequency distribution of the senses of words, particularly topical words, to depend on the genre and domain of the text under consideration. We present work on the use of a thesaurus acquired from raw textual corpora and the WordNet similarity package to find predominant noun senses automatically. The acquired predominant senses give a precision of 64% on the nouns of the SENSEVAL-2 English all-words task. This is a very promising result given that our method does not require any hand-tagged text, such as SemCor. Furthermore, we demonstrate that our method discovers appropriate predominant senses for words from two domain-specific corpora.

Subfields: Artificial Intelligence, Developmental and Educational Psychology

View Paper: https://www.semanticscholar.org/paper/1aac9a51700b4f548ed4d406d3987c8008876521

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 Developmental and Educational Psychology

Citations in the corpus

Word Sense Disambiguation with Distribution Estimation (2005, methodology): “A method of automatically ranking WordNet senses to determine the predominant, or most frequent sense, of a noun in the BNC corpus is presented in [McCarthy et al., 2004b].”
Corpus-based and Knowledge-based Measures of Text Semantic Similarity (2006, methodology): “The Wu and Palmer (Wu & Palmer 1994) similarity metric measures the depth of two given concepts in the Word- 1 This is similar to the methodology used by ( McCarthy et al. 2004 ) to find similarities between words and senses starting with a concept-to-concept similarity measure.”
A Context-Theoretic Framework for Compositionality in Distributional Semantics (2011, methodology): “These techniques have proved useful in many applications within computational linguistics and natural language processing (Grefenstette 1994; Schütze 1998; Bellegarda 2000; Choi, WiemerHastings, and Moore 2001; Lin 2003; McCarthy et al. 2004), arguably providing evidence that they capture something ”
How Dominant Is the Commonest Sense of a Word? (2004, background): “Recent work by McCarthy et al. [ 14 ] has taken this strategy a step further, exploring in detail how different word senses are commonest in different domains, and how NLP application performance can be improved by using this information.”
Measuring semantic similarity between words using web search engines (2007, background): “Disambiguating named entities is important in various applications such as information retrieval [9], social network extraction [19, 3, 4], Word Sense Disambiguation (WSD) [21], citation matching [11] and cross-document co-reference resolution [28, 10].”
Structural semantic interconnection: a knowledge-based approach to Word Sense Disambiguation (2005, background): “The first sense heuristic was a common choice for many systems to fulfill the requirement of high recall [30].”
An Experimental Study of Graph Connectivity for Unsupervised Word Sense Disambiguation (2010, methodology): “It is therefore interesting to see that our algorithm fares well in comparison with other more sophisticated methods that exploit syntactic information (e.g., [ 10 ], [72]) or additional resources (e.g., [68]).”
Measures of semantic similarity and relatedness in the biomedical domain (2007, background): “[11] carry out word sense disambiguation based on the idea that a word should be used in the sense that is most similar to or related to the sense of the words that surround it.”
Dependency-Based Construction of Semantic Space Models (2007, methodology): “Our following experiment applies the dependency space introduced in this article to word sense disambiguation (WSD), a task which has received much attention in NLP and is ultimately important for document understanding.”
On the Use of Automatically Acquired Examples for All-Nouns Word Sense Disambiguation (2008, methodology): “Regarding the improvement of SenseCorpus-U over the Automatic Ranking baseline (Au-99 toPS), the best results are obtained in the low-frequency range (0-20), when the baseline scores in the 50-60% F-score range.”
A Topic Model for Word Sense Disambiguation (2007, background): “Like McCarthy (2004), our unsupervised system acquires a single predominant sense for a domain based on a synthesis of information derived from a textual corpus, topics, and W ORD N ET -derived similarity, a probabilistic information content measure.”
Word Sense and Subjectivity (2006, methodology): “This idea relates to the unsupervised word sense ranking algorithm described in (McCarthy et al., 2004).”
Representing words as regions in vector space (2009, methodology): “In NLP, vector space models have featured most prominently in information retrieval (Manning et al., 2008), but have also been used for ontology learning (Lin, 1998; Snow et al., 2006; Gorman and Curran, 2006) and word sense-related tasks (McCarthy et al., 2004; Schütze, 1998).”
Personalizing PageRank for Word Sense Disambiguation (2009, methodology): “ric is used for calculating the relatedness among senses (Lesk, 1986; McCarthy et al., 2004).”
Unsupervised Visual Sense Disambiguation for Verbs using Multimodal Embeddings (2016, background): “Unsupervised WSD approaches often rely on distributional representations, computed over the target word and its context (Lin, 1997; McCarthy et al., 2004; Brody and Lapata, 2008).”
Deep Learning for Answer Sentence Selection (2014, background): “They have been proved to be successful in applications such as relation extraction [7] and word sense disambiguation [14].”
A Long Short-Term Memory Model for Answer Sentence Selection in Question Answering (2015, background): “…network-based distributed sentence modeling has been found successful in many natural language processing tasks such as word sense disambiguation (McCarthy et al., 2004), discourse parsing (Li et al., 2014), machine translation (Sutskever et al., 2014; Cho et al., 2014), and paraphrase detection…”