From: Natalia Levshina <natalevsgmail.com>
Subject: Quantitative Methods in Cognitive Semantics: Corpus-driven approaches
E-mail this message to a friend
Discuss this message
Announced at http://linguistlist.org/issues/21/21-4367.html
EDITORS: Glynn, Dylan and Fischer, KerstinTITLE: Quantitative Methods in Cognitive Semantics: Corpus-driven approachesSERIES TITLE: Cognitive Linguistics Research 46PUBLISHER: De Gruyter MoutonYEAR: 2010
Natalia Levshina, RU Quantitative Lexicology and Variational Linguistics,Department of Linguistics, K.U. Leuven
Created as a follow-up of a workshop at the International Cognitive LinguisticsConference 2007 in Krakow, ''Quantitative Methods in Cognitive Semantics:Corpus-driven approaches'' provides an insight into the main tendencies in thisdynamic research field. The book is primarily targeted at readers withbackground in Cognitive Linguistics and Cognitive/Construction Grammar --disciplines where the methodological standards are subject to considerabledebate. The volume provides general epistemological arguments in favour of theempirical approach to semantics, and demonstrates how this approach can work ina number of case studies. It also identifies some conceptual and practicalcaveats and pitfalls in this challenging enterprise, and charts a path forfuture research.
The volume opens with two introductions. The first one, written by Dylan Glynn,introduces the field of empirical Cognitive Semantics. Going back to the rootsof Cognitive Linguistics, Glynn shows that the empirical 'bomb' was already laidin the foundations of the discipline, and argues that the fundamental conceptsof entrenchment and categorization can be successfully operationalized in aquantitative way. His contribution includes a comprehensive yet concise surveyof the field (two tables on pp. 23-24) and an extensive list of references. Thesecond introduction, written by Kerstin Fisher, presents the structure and thecontents of the volume. It outlines the statistical methods employed in thestudies and discusses the semantic aspects that the contributors focus on. Theintroduction contains a short summary of each contribution and describes itsplace in the structure of the volume.
Section I ''Corpus Methods in Cognitive Semantics'' discusses the mainmethodological principles and challenges of contemporary corpus-based CognitiveSemantics, setting the stage for the following case studies.
The first contribution of the volume is Dirk Geeraerts's paper entitled ''Thedoctor and the semantician''. Beginning with a metaphorical comparison betweensemantic analysis and medical diagnostics, the paper explores the relationshipsbetween introspection and empirical methods in semantic research. Geeraertsargues that neither approach is sufficient on its own. On the one hand,introspection fails in such fundamental tasks as demarcating and observinglinguistic objects. On the other hand, Geeraerts warns against empiricalfetishism and demonstrates that a linguist's intuition is an important componentof the empirical cycle.
In his paper ''Balancing acts: Empirical pursuits in Cognitive Linguistics'' JohnNewman shares his concerns about the current practices in corpus-drivensemantics. He points out that communicative context is not taken into accountsufficiently, and that the fundamental genre of discourse -- spontaneousface-to-face conversation -- is underrepresented in corpus studies. Newman alsodiscusses hurdles and caveats of corpus tagging and states that inflected formsare commonly overlooked in favour of lemmas. He concludes with a plea for abalance between various kinds of evidence and different statistical methods.
Hans-Jörg Schmid asks a fundamental question: ''Does frequency in textinstantiate entrenchment in the cognitive system?'' in his contribution of thesame title. There is no clear answer to this question yet, which leads to arange of methodological problems with frequency-based corpus-driven measures ofmutual attraction between constructions and lexemes. Comparing his ownattraction-reliance method and Collostructional Analysis (Stefanowitsch andGries 2003), Schmid shows that the measure of attraction used in the latter maymask different distributional relationships. Another major issue of concern isthe relationship between the frequency of a lexeme in a corpus and its relativefrequency in a construction. The author believes that resolving these andrelated issues is possible through integration of corpus-based and experimentalevidence.
Section II ''Advancing the science: Theoretical questions'' contains four papersthat offer empirical approaches to such core concepts in Cognitive Semantics asforce dynamics, mood, aspect and semantic productivity.
In his paper ''The aspectual coercion of the English durative adverbial,'' StefanFuhs discusses the durative adverbial construction (e.g. ''work for threehours''), which is frequently used in aspectual studies as a test for(a)telicity. With the help of Collostructional Analysis of the verbal slot, heshows that the construction also attracts a significant number of telic verbs.He suggests that the construction can coerce a telic verb into an atelicmeaning, thus shifting the aspectual profile of the verb. This finding hasimportant theoretical consequences, as it questions the existence of inherentlexical aspect.
Martin Hilpert's chapter entitled ''The force dynamics of English complementclauses'' focuses on English gerund clauses with an infinitive complement, as in''Learning to read is fun''. Hilpert performs a collexeme analysis to show thatthe construction has a range of force-dynamic meanings (cf. Talmy 2000), such asan attempt or obligation. The study thus indicates that force-dynamic semanticscan be expressed by complement clauses, besides the grammatical domains ofcausation and modality. The author also demonstrates that the results of hiscorpus-based analyses align with intuitive grammaticality judgments.
In her paper ''Accounting for the role of situation in language use in aCognitive Semantic representation of sentence mood,'' Kerstin Fischer studies howthe differences in speakers' construal of one and the same situation influencetheir choice of sentence mood. Based on a unique corpus of human-robotinteraction, the study analyzes correlations between the use of declarative,imperative and interrogative sentences, on the one hand, and the frequencies ofsuch discourse elements as politeness formulas and dialogue openings, on theother hand. The correlated features are interpreted as evidence of differentconstruals of the human-robot communication situation. The author concludes withan interpretation of the findings in terms of Embodied Construction Grammar.
Arne Zeschel's contribution ''Exemplars and analogy: Semantic extension inconstructional networks'' is a corpus-based study of polysemy of the Germanadjective ''tief'' (''deep''). Trying to find out how speakers derive a schema fromthe exemplars of a construction, Zeschel formulates the following hypothesis:the distribution of novel adjective-noun combinations correlates with the numberof established adjective-noun combinations in different regions of the semanticmap. This hypothesis is tested and confirmed on three different levels ofsemantic specificity.
Section III, entitled ''Advancing the scene: Methodological questions,'' focusesmainly on specific methodological problems of empirical Cognitive Semantics,although the issues discussed here are also theoretically important.
This section begins with Stefanie Wulff's contribution ''Marryingcognitive-linguistic and corpus-based methods: On the compositionality ofEnglish V NP idioms''. The paper is a case study of V NP-constructions inEnglish. The author proposes an original corpus-linguistic measure ofcompositionality, which takes into account the semantic contributions of theverb and the noun in the constructional meaning. The measure is based on thedistribution of collocates of the verb and the noun separately and together. Theresults reveal a continuum of idiomaticity from the highly idiomatic ''make DETheadway'' to the fully compositional ''write DET letter'' (where DET stands forDeterminer) with metaphorical and quasi-metaphorical extensions in between.
The chapter written by Dylan Glynn is entitled ''Testing the hypothesis.Objectivity and verification in usage-based Cognitive Semantics''. He discussesthe problem of operationalizing abstract semantic features in a corpus-basedanalysis and illustrates it with a case study of the semantics of ''bother'' .Using correspondence analysis and logistic regression, Glynn identifies threesenses of ''bother'', which correspond to specific constructional patterns,forming thus a small onomasiological field within the lexeme. This findingallows him to question the theoretical validity of the traditional distinctionbetween semasiology and onomasiology (e.g. Geeraerts, Grondelaers and Bakema 1994).
In his contribution ''Beyond the dative alternation: The semantics of the Dutchaan-Dative'' Timothy Colleman argues that near-synonymous constructions('alternations') should be studied in their own right, not (only) in terms oftheir distinctive features (cf. Goldberg 2002). He performs an independentsemantic analysis of the aan-Dative in Dutch, which has been traditionallydescribed in contrast with the ditransitive construction. Using a collexemeanalysis of newspaper data, Colleman interprets the broad semantic range of theconstruction in accordance with the multidimensional approach to polysemyproposed in Geeraerts (1998).
Dagmar Divjak's chapter ''Corpus-based evidence for an idiosyncraticaspect-modality relation in Russian'' raises the question how two fundamentalconceptual domains of aspect and modality interact in combinations of a modaladverbial predicate and an (im)perfective infinitive. The author uses logisticregression with mixed effects applied to data from a literary corpus to find outthat dynamic modality is usually associated with the perfective infinitive,whereas the deontic modal predicates co-occur with the imperfective infinitive.She argues that these associations reflect the similarities of aspect andmodality on a deep conceptual level, which involves ''knowledge about the way inwhich time affects situations and the written and unwritten code that rules ourlives'' (p. 324).
The final Section IV, ''Towards an empirical Cognitive Semantics,'' discusses theperspectives of the approach and deals with some of the criticisms leveled atcorpus-driven Cognitive Semantics by opponents.
The highly polemic contribution by Stefan Gries and Dagmar Divjak entitled''Quantitative approaches in usage-based Cognitive Semantics: Myths, erroneousassumptions and a proposal'' addresses the most vociferous criticisms againstcorpus-based Cognitive Linguistics (e.g. Raukko 2003, Talmy 2000), claiming thatthey are either truisms, misunderstandings or misrepresentations. Thecontributors then propose their own Behavioral Profiles approach, amulti-purpose corpus-driven bottom-up model of semantics, and show that theresults of this approach is supported by a range of validation techniques andconverging experimental evidence.
The final paper in the volume is Anatol Stefanowitsch's ''Empirical CognitiveSemantics: Some thoughts''. He discusses four types (operationalizations) ofmeaning, each with its own methods and techniques, and proposes the steps thatthe research field should take to become a full-fledged scientific discipline,instead of just an ''exercise in speculative psychology'' (p. 374). These stepsinvolve adopting the protocols of empirical research, redefining all theoreticalconcepts in an empirically operationalizable way, and -- the most radical step-- giving up the concepts that cannot be operationalized.
Overall, this volume is an important contribution to the development ofempirical Cognitive Semantics. This collection of high-quality papers providesthe reader with an insight into the most important empirical approaches incorpus-driven semantic research. Originating from a conference event, the bookraises the questions that many empirically minded cognitive linguists areconcerned with nowadays. Inevitably, it also mirrors some of the biases incontemporary empirical Cognitive Semantics: a strong theoretical focus onEnglish data and constructional semantics, with Collostructional Analysis asarguably the most popular tool. Probably because it has taken the volume aboutthree years to appear, the book does not reflect a more recent trend ofintegrating semantic and contextual variation (e.g. Geeraerts, Kristiansen andPeirsman 2010), although Glynn does mention this tendency in his Introduction. Aminor criticism concerns the composition of the volume: the reviewer has foundat least four introductions to Collostructional Analysis in different papers. Apossible solution might have been to present the common quantitative methodsonly once in an introduction or appendix.
From the stylistic point of view, the book contains very different voices, whichmakes it an exciting reading -- from the metaphoricity of Geeraerts's paper toGries and Divjak's passionate polemic, and a glimpse into the future of thefield in Stefanowitsch's chapter. This diversity in combination with the highprofessional level of the contributions makes one believe that the volume willreach a broad audience and will have the desired impact on the development ofCognitive Semantics.
Geeraerts, Dirk. 1998. The semantic structure of the indirect object in Dutch.In Willy Van Langendonck and William Van Belle (Eds.), The Dative. Vol. 2.Theoretical and contrastive studies, 185-210. Amsterdam: John Benjamins.
Geeraerts, Dirk, Stefan Grondelaers and Peter Bakema. 1994. The structure oflexical variation. Meaning, naming and context. Berlin: Mouton de Gruyter.
Geeraerts, Dirk, Gitte Kristiansen and Yves Peirsman (eds.). 2010. Advances inCognitive Sociolinguistics. Berlin: Mouton de Gruyter.
Goldberg, Adele E. 2002. Surface Generalizations: An alternative toalternations. Cognitive Linguistics 13. 327-356.
Raukko, Jarno. 2003. Polysemy as flexible meaning: Experiments with English getand Finnish pitää. In Brigitte Nerlich et al. (eds.), Polysemy: flexiblepatterns in mind and language, 161-193. Berlin: Mouton de Gruyter.
Stefanowitsch, Anatol and Stefan Th. Gries. 2003. Collostructions: Investigatingthe interaction of words and constructions. International Journal of CorpusLinguistics 8(2). 209-243.
Talmy, Leonard. 2000. Toward a Cognitive Semantics. Cambridge: MIT Press.
ABOUT THE REVIEWER
Natalia Levshina is finishing her PhD at the University of Leuven. Her main research interests are in constructional and lexical semantics. She is particularly interested in quantitative corpus-driven methods of modeling semantic and contextual variation with the help of multivariate statistical techniques, and in integration of exemplar and prototype theories of categorization with corpus linguistics.
Page Updated: 26-Jul-2011