D. Glynn – Corpus Methods for Semantics

Автор: D. Glynn
Название книги: Corpus Methods for Semantics
Формат: PDF
Жанр: Лингвистика
Страницы: 554
Качество: Изначально компьютерное, E-book

This volume seeks to advance and popularise the use of corpus-driven quantitative methods in the study of semantics. The first part presents state-of-the-art research in polysemy and synonymy from a Cognitive Linguistic perspective. The second part presents and explains in a didactic manner each of the statistical techniques used in the first part of the volume. A handbook both for linguists working with statistics in corpus research and for linguists in the fields of polysemy and synonymy.

It could be argued that Cognitive Linguistics is undergoing a paradigm shift. Originally,
the field sought to show the inadequacies of earlier models of language and the
theories of linguistic structure based upon them. Today, the emphasis has changed
to testing the various theories about how language works (Geeraerts 2006; Gries and
Stefanowitsch 2006; Stefanowitsch and Gries 2006; Gonzalez-Marquez et al. 2008;
Glynn and Fischer 2010). This has brought analytical methods, based on observable
and quantifiable data, to the fore. In the light of these developments, this volume systematises,
reviews, and promotes a range of research techniques and theoretical perspectives
that currently inform work across the field of linguistics, with a particular
focus on Cognitive Semantics. More precisely, the aim of this book is twofold:
i. Didactic: To broaden the understanding and application of the state-of-the-art
corpus linguistic techniques for the study of conceptual structure in Cognitive
Semantics.
ii. Scientific: To advance the state-of-the-art of those techniques through a collection
of studies applied to the description of the conceptual structures of polysemy
and synonymy.
This publication grew out of the belief that there exists a strong desire in the research
community to understand and learn how quantitative corpus methods work and how
to apply them to research questions that are basic to the cognitive project. Instead of a rift
between linguists using corpus data and those using traditional introspective analysis,
constructive communication between the methodologies should be encouraged. Both
the descriptive research and the explanations of the statistical techniques included in
this book seek to promote such communication. The chapters that describe the statistical
techniques are written to help linguists using traditional methods both understand
how these new methods work and how to apply them. The research chapters, in turn,
showcase the methods described. Their aim is not only to advance corpus-driven quantitative
research in Cognitive Semantics, but also to promote the possibilities that these
methodologies offer. Observational data and quantitative corpus-driven methods
cannot inform all research questions. However, it is hoped that this volume will advance
the current state-of-the-art in their use as well as promote their application in
the broader linguistic research community

The book divides into two sections. The first section begins with eleven chapters, arranged
according to their object of study. These chapters begin with an overview of
the field in “Polysemy and synonymy: Corpus method and cognitive theory” (Glynn).
This chapter includes the analytical justification for approaching both lexis and morpho-
syntax in terms of polysemy and synonymy as well as a justification of extending
the traditional uses of the terms to cover any variation or similarity in use. The analytical
chapters begin with morpho-syntactic polysemy, move to lexical polysemy, then
on to lexical synonymy, and finally turn to morpho-syntactic synonymy.
Beginning with research on the polysemy of morpho-syntactic semantics, the first
descriptive chapter, “Competing ‘transfer’ constructions in Dutch” (Delorge, Plevoets,
and Colleman), considers the polysemy of a morpheme-based construction. The ontprefix
in Dutch combines with a range of verbs to express dispossession. Using correspondence
analysis, the study seeks to capture the lexical semantic morpho-syntactic
interplay associated with the construction. The next chapter, “Rethinking constructional
polysemy” (Perek), also examines a grammatical construction. The syntactically
encoded conative construction in English combines with a range of lexemes.
Through the application of collostructional analysis, the author attempts the task of
teasing out and identifying the semantic variation associated with the construction.
Turning to lexical semantics, “Quantifying polysemy in cognitive sociolinguistics”
(Robinson) examines the usage of polysemous adjectives in a community of
speakers from South Yorkshire, UK. The study applies cluster analysis, logistic regression,
and decision tree analysis in order to examine the extent to which individual
conceptualisations are non-random and can be related to the socio-demographic
characteristics of the speaker. “The many uses of run” (Glynn) is a repeat analysis
of Gries’ (2006) study. Employing a combination of cluster analysis, correspondence
analysis and logistic regression, it confirms Gries’ findings but argues that sociolinguistic
dimensions should be included in the study of polysemy.
Remaining with lexical semantics, but focusing on near-synonymy, “Visualizing
distances in a set of near-synonyms” (Desagulier) examines rather, quite, fairly,
and pretty in English, combining collostructional analysis and multivariate statistics
such as correspondence analysis and cluster analysis. “The uses of may and can in
French-English interlanguage” (Deshors and Gries) treats a lexical alternation in first
language and second language use. With the use of cluster analysis and logistic regression,
the authors seek to identify not only the relationship in use between may and
can, but to compare this with French native speakers using English.
Moving towards morpho-syntactic semantics, “Dutch causative constructions”
(Levshina, Geeraerts, and Speelman) examines a lexeme-based grammatical construction
alternation. Focusing on the expression of causation in Dutch, the study employs
logistic regression analysis to determine both the semantic and extralinguistic factors that determine the choice and difference in conceptualisation between the two constructions.
The next study, “The semasiological structure of Polish myśleć ‘to think’”
(Fabiszak, Hebda, Kokorniak, and Krawczak) continues to move from lexical to syntactic
semantics with an analysis of the near-synonymy of a set of prefix-verb combinations.
The study combines introspective methods and usage-feature analysis, examined
with cluster analysis, correspondence analysis and logistic regression.
“A multifactorial analysis of grammatical synonymy” (Klavan) studies a lexical-
morphological alternation. Employing logistic regression, the study attempts to
determine the conceptual differences that motivate speakers’ choice of a preposition
over a grammatical case to express the spatial relation of on in Estonian. “A diachronic
corpus-based multivariate analysis of ‘I think that’ vs. ‘I think zero’” (Shank, Plevoets,
and Cuyckens) is an analysis of well-known complementiser alternation in English.
Logistic regression is used to test a wide range of language factors proposed in the
literature to motivate the omission of the complementiser.
The second section consists of seven chapters that explain some of the tools and
methods for quantitative corpus-driven research. It is designed to introduce the application
and interpretation of the statistical methods used in the first section for
researchers completely new to the field. It also serves as a ‘cookbook’, or is a quick reference,
for intermediate users of the statistical techniques and the programming environment
R. The first chapter, “Techniques and tools: Corpus methods and statistics
for semantics” (Glynn), is an overview of the field. It examines two corpus methods
that are commonly used in Cognitive Semantics and summarises many of statistical
techniques currently used in the field.
The second chapter introduces the statistical environment of R, used throughout
the book (van de Weijer and Glynn). Readers with no experience in R will find this
chapter useful when applying the techniques described in the previous chapters to
data analysis. The following chapter, “Frequency tables: Tests, effect sizes, and explorations”
(Gries), covers many of the essential and basic analytical concepts and how
to apply them in R. Building on the statistical basics, the next chapter, “Collostructional
analysis: Measuring associations between constructions and lexical elements”
(Hilpert), explains the application of collostructional analysis, one of the most popular
quantitative techniques in Cognitive Linguistics. This family of techniques are
used to quantify the degree of association between linguistic forms.
The next three chapters each consider a different multivariate statistical technique.
The three techniques in question have proven popular in recent Cognitive Linguistic
research. The chapter “Cluster analysis: Finding structure in linguistic data”
(Divjak and Fieller), focuses on a method for sorting a given set of phenomena, such
as lexemes, constructions, or senses, into categories of similar and dissimilar, relative
to some other set(s) of linguistic phenomena such as meanings, argument types, case
marking, and so forth. This is followed by the chapter “Correspondence analysis: Exploring
data and identifying patterns” (Glynn), which considers a technique similar to cluster analysis, but one that looks for correlations between different phenomena
rather than categorising them. It is useful for identifying structure in complex multidimensional
data. The third chapter on multivariate techniques, “Logistic regression:
A confirmatory technique for variant comparison in corpus linguistics” (Speelman),
considers an advanced form of statistical modelling. Logistic regression is a powerful
and popular tool in the social sciences, including Cognitive Linguistics. As a confirmatory
technique, regression analysis represents a level of statistical analysis that is
more complex than the previous techniques covered. This chapter charts the basics of
its application, interpretation and verification.
Where the first section represents a broad, yet coherent, picture of the cutting
edge in the application of these techniques, the second section seeks to offer an introduction
to the different statistical techniques employed in corpus-driven semantics.
Focusing on the study of polysemy and synonymy of both lexical morpho-syntactic
forms, these empirical analyses represent the vanguard of corpus-driven Cognitive
Linguistics.