Learning Word Sense Distributions, Detecting Unattested Senses and Identifying Novel Senses Using Topic Models

Abstract

Unsupervised word sense disambiguation (WSD) methods are an attractive approach to all-words WSD due to their non-reliance on expensive annotated data. Unsupervised estimates of sense frequency have been shown to be very useful for WSD due to the skewed nature of word sense distributions. This paper presents a fully unsupervised topic modelling-based approach to sense frequency estimation, which is highly portable to different corpora and sense inventories, in being applicable to any part of speech, and not requiring a hierarchical sense inventory, parsing or parallel text. We demonstrate the effectiveness of the method over the tasks of predominant sense learning and sense distribution acquisition, and also the novel tasks of detecting senses which aren't attested in the corpus, and identifying novel senses in the corpus which aren't captured in the sense inventory.

Improving selection of synsets from WordNet for domain-specific word sense disambiguation

“…The results were compared against those reported in the literature (Lau et al, 2014;Navigli et al, 2011;Koeling et al, 2005). Results of the proposed approach represent an improvement over the results presented by Lau et al, (2014) who used topic models, Navigli et al (2011) who used the Semantic Model Vector, who used the Personalized Page Rank algorithm, and the results given by Koeling et al (2005) who used predominant sense acquisition from a domain-specific corpus. The obtained results by the proposed approach on the Sports and Finance domains outperform those reported in the literature.…”

“…The developed prototype has as drawback that by using the Web as linguistic resource, sometimes the generated auxiliary corpus could be poor; this process requires longer time for searching relevant webpages. (Lau et al, 2014) 42.2 -55.5 -37.6 -Semantic model vector (Navigli et al, 2011) -52.7 -58.2 --Predominant sense (Koeling et al, 2005) 49. …”

“…Since each ambiguous word has several senses the 100% of agreement is never achieved. These corpora have been used in previous works in the state of the art (Koeling et al, 2005;Navigli et al, 2011, Lau et al, 2014.…”

“…Knopp et al (2013) proposed to use topic models as a way to estimate the distribution of word sense in text, then they used topic-word distributions as a way to derivate a semantic representation of ambiguous words in context, theses distributions were the input to a k-means algorithm to identify senses of words. Lau et al (2014) proposed an unsupervised topic modelling-based approach. They considered each topic as a sense of the target lemma.…”

A comparison of homonym meaning frequency estimates derived from movie and television subtitles, free association, and explicit ratings

“…In computational linguistics, the similar task of word sense 1 disambiguation-automatically identifying which meaning of an ambiguous word has been evoked in a particular context-has been an important and long-studied theme (Agirre & Edmonds, 2007;Bartunov, Kondrashkin, Osokin, & Vetrov, 2016;Lefever & Hoste, 2010;Li & Jurafsky, 2015). Recent work in this field has also focused on developing unsupervised methods for determining relative meaning frequencies, or sense distributions for ambiguous words, including but not limited to homonyms (e.g., Bennett et al, 2016;Lau et al, 2014). Bringing these psycholinguistic and computational linguistic strands of work together, it may be possible to partially or fully automate the estimation of homonym meaning frequencies both in natural language and in the behavioral responses derived from empirical studies of semantic ambiguity.…”