Automatic extraction of semantic relations from specialized corpora

Abstract

In this paper we address the problem of discovering word semantic similarities via statistical processing of text corpora. We propose a knowledge-poor method that exploits the sentencial context of words for extracting similarity relations between them as well as semantic in nature word clusters. The approach aims at full portability across domains and languages and therefore is based on minimal resources.

Wolves in sheep’s clothing? Rhetorical strategies in the response narratives of Big Oil to IPCC evidence of climate change

“…Using the LancsBox corpus linguistic software suite (Brezina et al, 2020), we used the MI3 statistical measure of associative strength. Known also as either pointwise mutual information cubed, or log-frequency biased mutual dependency (Thanopoulos et al, 2002), the MI3 is a bidirectional, contingency-based measure of association. It favours rare or unique but statistically significant collocations between the key (search) word and other terms in the corpus that cannot be explained by reference to the probability of random collocations alone.…”

Insights Into Automatic Extraction of Metaphorical Collocations

“…The mutual dependency (MD) proposed by Thanopoulos, Fakotakis and Kokkinakis (2002) is based on the idea that dependency can be identified by subtracting information that the whole event carries from the Pointwise Mutual Information (PMI) score. Walker (2011) extracts the most frequent collocates using t-score with respect to position and considering raw frequency.…”

“…There is a line of research that is characterised by attempting to compare existing methods, and not combining or developing new methods (e.g., Krenn and Evert 2001;Thanopoulos et al 2002;Pearce 2002;Pecina 2005). Pecina (2005) presents the most extensive empirical evaluation which includes 84 automatic collocation extraction methods.…”

“…They do not usually retain the meaning of some or even all of their components across languages, making them somewhat difficult for non-native speakers to learn. They are domain-dependent and language-dependent and differ in terms of length, syntactic patterns, and spacing between constituents (Thanopoulos et al 2002).…”

“…Strakatova et al (2020) differentiate between semantically opaque idiomatic expressions and collocations that are not fully lexicalized. Some other authors such as Thanopoulos, Fakotakis and Kokkinakis (2002) place collocations in the subcategory of named entities.…”