AFLiCo Lecture Series 2020

C'est la nouveauté de l'année 2020 : l'Association française de linguistique cognitive (AFLiCo) vous propose une série de conférences invitées. Cette proposition fait suite à nos traditionnelles journées d'études (JET AFLiCo) qui ont très bien fonctionné depuis 10 ans, et ont permis des échanges féconds sur la linguistique cognitive et l'ensemble des approches basées sur l'usage (voir par ex. le numéro 19 de CogniTextes qui est issu de notre cinquième JET). La sixième édition devait avoir lieu à Grenoble à l'automne 2020, mais la crise sanitaire nous a empêché de nous retrouver, et nous avons décidé de proposer un nouveau format. La série de conférences invitées a eu lieu en ligne du 14 au 18 décembre 2020 et a fait aussi l'objet d'une publication dans le Volume 21 de CogniTextes.

Programme

Cliquez sur les liens ci-dessous pour accéder directement au résumé de chaque intervention. Les horaires sont tous à l'heure de Paris.

Lundi 14 décembre 2020: Kari Sullivan, 12h30-14h30 (CET)

Abstract artworks 'speak' to fewer people and have less to 'say' than figurative works

When an artwork interests you, the work metaphorically 'speaks' to you. But what determines whether it 'speaks' and what it 'says'? In this talk, I'll discuss two studies showing that artworks' subject matter affects how they 'speak'. Both studies look at corpora of artists' statements, and analyse how artists metaphorically describe their artworks as speaking, telling stories, screaming, or otherwise vocalising.

Both studies point to differences between purely abstract works and figurative works (i.e. those that depict people, objects or landscapes). The first study finds that abstract artworks usually 'speak' to their creators, whereas figurative works mostly 'speak' to their viewing audience (Sullivan 2006, 2009). For instance, abstract artists often claim to engage in dialogue with their canvases. Figurative artists more often intend their work to communicate with the viewer.

The second study finds that abstract artworks not only 'speak' to fewer people, but are also less likely to 'speak' in directly presented speech (Sullivan 2016). For example, a figurative painting might say, 'Look at me!' in directly presented speech, whereas an abstract work is more likely to scream for attention without any direct speech attributed to the artwork.


Mardi 15 décembre: Stefan Th. Gries (UC Santa Barbara & JLU Giessen), 17h30-19h30 (CET)

Increasing (quantitative) precision: blend production, contexts of alternations, and corpus-linguistic association

In this talk I will discuss three different case studies, all of which are concerned with increasing the degree of precision — especially quantitative precision — of previous work in well-researched areas of linguistic study.

Part 1 is concerned with morphological blends (e.g. breakfast x lunch → brunch). While much traditional research has concluded that blends are formed largely arbitrary, research over the last 20 years or so has discovered a variety of probabilistic patterns governing the selection of words to blend and the way they are merged into a blend. However,  much of this research — including my own — has been based on what are ultimately convenience samples: collections of blends encountered in 'the wild', which may distort the frequencies with which certain patterns are observed. To test the observational data's validity, Stefanie Wulff and I did a series of blend production experiments under controlled conditions and I will report a few very small case studies designed to determine whether certain observational results are confirmed or not.

Part 2 is a very exploratory and tentative kind of suggestion for the corpus-based analysis of the lexical context of syntactic alternations. Studies of alternations/choices in particular in corpus linguistics have become increasingly sophisticated in terms of the statistical methods they employ and the ever larger number of predictors they involve many different levels of linguistic analysis — phonology, morphosyntax, semantics, pragmatics/discoursal, textual, psycholinguistic, sociolinguistic, and others. These predictors are usually contextual in nature, meaning they characterize the context of the choice the language user needs to make or has just made. However, one aspect of the context seems to be crucially underutilized when it comes to modeling speakers' choices: the actual lexical context. In this part, I use recent work in computational psycholinguistics to (i) define a lexical-distribution prototype of each of the (typically, but not necessarily, two) alternants of an alternation and (ii) compute the degree to which each instance of the alternation in question diverges from each of the prototypes. Then, (iii) the values that all choices score on the divergences from each of the prototypes are entered as predictors to all others in statistical models to, minimally, serve as a variable that controls for whatever information is contained in the lexical context of an instance of speaker's choice. I exemplify the approach and its sometimes amazing predictive power on the basis of a choice between near synonyms and two morphosyntactic alternations (preposition stranding vs. pied-piping and of- vs. s-genitives).

Part 3 discusses a variety of potential shortcomings of most of the most widely-used association measures as used in collocational/collostructional research. To address these shortcomings, I then discuss a research program called tupleization, an approach that does away with the usual kinds of information conflation by keeping relevant corpus-linguistic dimensions of information — e.g., frequency, association/contingency, dispersion, entropy, etc. — separate and analyzing them in a multidimensional way; I conclude with pointers towards how these dimensions could, if deemed absolutely necessary, be conflated for the simplest kinds of of rankings as well as strategies for future research.


Mercredi 16 décembre: Reyes Llopis-García (Columbia University), 17h30-19h30 (CET)

No Functions... Constructions! CxG and Cognitive Linguistics: Powerful Allies for User friendly L2 Pedagogy

Most L2 teaching methodologies and textbooks, especially where explicit instruction of grammar is concerned, have remained largely unchanged since the early 1980s (Tyler 2012, Larsen-Freeman 2015), with their "scope and sequence" vertebrated into notions (i.e. telling stories in the pastor giving advice) and functions that match them (i.e. the preterit and the imperfect or using the conditional and the imperative, etc.). Notions then provide the lexical units necessary to create language around them, while functions teach the grammatical items that structure their content. And while this is the most widely accepted manner of working in the L2 classroom, it is by no means the best or the only way. In fact, usage-based approaches to L2 teaching and learning, like Applied Cognitive Linguistics (ACL) and Sociocultural Theory have been gaining ground and popularity in professional development and academic training for language instructors, and the last few years have seen the first footprints of ACL in textbooks and grammar manuals from the editorial world, as well as in the way some instructors are approaching the teaching of their target language: prototypes and SCOBAS, metaphorical competence, motion and embodiment, or perspective are slowly becoming more frequent and more appreciated in the L2 classroom by teachers and students alike because they make grammar and lexis more accessible and comprehensible for both parties (Llopis-García, Real Espinosa & Ruiz Campillo 2012, Castañeda Castro 2014, Ibarretxe-Antuñano, Cadierno & Castañeda Castro 2019, Llopis-García & Hijazo-Gascón 2019).

Enter Constructions, and the grammar-lexis continuum acquires new opportunities for further accessibility to not only meaningful form-meaning pairings, but also to a much-needed rhyme and reason in the combination and behavior of linguistic structures. There is enormous potential to CxG in the L2 language classroom, as evidenced both by a growing body of research (Holme 2010, De Knop & Gilquin 2016, Gras 2017, Ibarretxe-Antuñano & Cheikh-Khamis 2019, among many others), and an increasing interest from the teaching community in learning more about usage-based approaches to L2.This talk aims to contribute insights to CxG for the language classroom by exemplifying how constructions may be taught (with a focus on Spanish/L2, since specific teaching materials will be shown), by arguing for best practices for their teachability and their pedagogical inclusion, as well as strongly supporting the presence of CxG approaches from the very early stages of the elementary levels, because their two-fold stability and versatility is very attractive for beginners as they learn to navigate the structure of the new linguistic system. Finally, a case will be made for equal-opportunity collaborations between linguists and language instructors, since the knowledge of the former cannot come to fruitful research success without the input and know-how of the latter.

Castañeda Castro, A. (Ed.) 2014. Enseñanza de Gramática Avanzada de ELE. Madrid: SGEL.
DeKnop, S. & Guilquin, G. (Eds.) 2016. Applied Construction Grammar. Boston: DeGruyter.
Gras, P. 2017. Gramática de Construcciones para Profesores de ELE. Lecture at the Universitat de Barcelona. December 20th. Retrieved online from [https://www.youtube.com/watch?v=57FwAxyOmr0&list=WL&index=4]


Jeudi 17 décembre: Florent Perek (University of Birmingham), 17h30-19h30 (CET)

Construction Grammar in action: The English Constructicon project

In a radical departure from traditional and formal approaches to grammatical description, construction grammar takes the view that grammar is best described as a network of form-meaning pairs at various levels of generality. This approach has been very successful in various areas of research, including for instance language acquisition (Tomasello 2003, Goldberg 2019) and language change (Traugott & Trousdale 2013), and it has also made its way into more applied fields, such as language pedagogy (De Knop & Gilquin 2016) or natural language processing (Steels 2011).

Early examples of constructions, such as the ditransitive construction, the caused-motion construction, or the way-construction (Goldberg 1995), have been offered as compelling illustrations of the central tenets of the theory. However, the range of constructions documented by construction grammar studies is still relatively limited, even in such an over-studied language as English. Many construction grammarians have tended to favour a "butterfly-collecting" approach, focussing primarily on patterns with unusual semantic and grammatical properties, but they have comparatively dealt less with common, mundane, and more predictable constructions, and it remains to be seen to what extent these constructions can equally be described in terms of form-meaning pairs. This "butterfly-collecting" attitude, and more generally the lack of larger-scale descriptive work, presents construction grammar with two problems. First, the descriptive bias towards idiosyncratic patterns has left the approach open to criticism, and has led it to be mostly perceived as primarily geared towards idioms, or the 'periphery' of grammar, especially from outside cognitive linguistics, with the 'core' grammar of common and regular patterns still seen as the province of formal approaches. Second, the lack of empirical coverage limits the applicability of the approach in applied linguistics, in particular for teaching purposes: if constructions are to be adopted more widely as a teaching approach, teachers and learners need descriptions of what constructions there are in a given language, especially the constructions that are most useful to language users.

Against this backdrop, in this talk I report on some early work on our English Constructicon project, one of the several constructicography projects currently going on to address this gap and provide large-scale descriptive research on constructions in various languages (Lyngfelt et al. 2018). Drawing on previous work from corpus linguistics and lexical semantics, our ultimate goal is to build a comprehensive inventory of grammatical constructions of the English language, following the principles and network approach of construction grammar (Perek & Pattern 2019). In the first phase of the project, we focus on constructions of the verb (aka argument structure constructions), but in the future we also aim to cover noun and adjective constructions. The English Constructicon is based on the COBUILD Grammar Patterns series (Francis et al. 1996, 1998), which provides an exhaustive list of English complementation patterns and the lexical items occurring in them. We follow a radically bottom-up approach in defining constructions: we first start from the entire list of verbs in a given pattern, and we use lexical semantic information from the FrameNet database (frames and frame-to-frame relations) and from COBUILD itself (verb groups) to posit narrow generalisations over verb meanings (or more precisely, the meaning these verbs take in the construction). We then use these groups to posit generalisations at higher levels, to arrive at a full inheritance hierarchy of constructions for each pattern.

Using the "V that", "V at n", "V n with n", and "V n n" patterns as examples, I illustrate how the English Constructicon data validate the constructional approach for more common and regular patterns than those typically described, while also capturing some of the key insights of the theory. First, there are intermediate levels of description in the inheritance hierarchy that allow us to strike a balance between the lowest level of the lexical item, which has a high degree of precision and detail but no predictive power, and the highest level of the construction as a whole, which is usually too broad to made reliable predictions about the distribution of the construction. Second, the constructional hierarchy can also capture idiosyncratic behaviour at any level of generality, both in form and in meaning. In particular, this allows us to describe as constructions patterns that would otherwise be treated as idioms or collocations and listed separately from the grammatical description (e.g. deal someone cards, bid someone farewell, in the "V n n" construction's hierarchy). These and other features make the constructional approach extremely attractive for the description of any kind of pattern, which has far-reaching implications for applied linguistics, especially in language teaching.

De Knop, S. & G. Gilquin, G. (eds.) (2016). Applied construction grammar. Berlin: Walter de Gruyter.
Francis, G., Hunston, S. & Manning, E. (1996). Collins COBUILD Grammar Patterns 1: Verbs. London: HarperCollins.
Francis, G., Hunston, S. & Manning, E. (1998). Collins COBUILD Grammar Patterns 2: Nouns and Adjectives. London: HarperCollins.
Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press.
Goldberg, A. E. (2019). Explain Me This. Princeton: Princeton University Press.
Lyngfelt, B., Borin, L., Ohara, K. & Torrent, T. (eds.) (2018). Constructicography: Constructicon development across languages. Amsterdam: Benjamins.
Perek, F. & Patten, A. (2019). Towards an English Constructicon using patterns and frames. International Journal of Corpus Linguistics, 24(3), 354–384.
Steels, L. (2011). Design patterns in Fluid Construction Grammar. Amsterdam: Benjamins.
Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. Cambridge, MA: Harvard University Press.
Traugott, E. C. & Trousdale, G. (2013). Constructionalization and Constructional Changes. Oxford: Oxford University Press.


Vendredi 18 décembre: Lauren Fonteyn (Leiden University Centre for Linguistics), 17h30-19h30 (CET)

Prepositional Polysemy through the lens of contextualized word embeddings

In recent years, contextualized embeddings generated by neural language models have grown extremely popular in Machine Learning community (e.g. Baroni, Dinu & Kruszewski 2014), but linguists seem generally more wary to use them. In this talk, I would like to highlight why it may be worth exploring to what extent linguists can (i) employ these models can be employed as tools, and (ii) help reveal what sort of information these models capture.

The first part of this talk discusses the practical application of contextualized embeddings. Focusing on embeddings created by the Bidirectional Encoder Representations from Transformer model, also known as 'BERT' (Devlin et al. 2019), I hope to demonstrate how contextualized embeddings can help counter two types of retrieval inefficiency scenarios that may arise with purely form-based corpus queries. In the first scenario, the formal query yields a large number of hits, which contain a reasonable number of relevant examples that can be labeled and used as input for a sense disambiguation classifier. In the second scenario, the contextualized embeddings of exemplary tokens are used to retrieve more relevant examples in a large, unlabeled dataset. As a practical case study, I will focus on the English preposition into (e.g. She got into her car / I'm so into you).

Subsequently, I will briefly turn to the question of whether these models can be employed as analytical tools to study meaning. In the second part (Part II) of this talk, I will focus on the principled polysemy model of the English preposition over as proposed by Tyler & Evans (2001) to investigate whether the sense network that emerges from this theoretical model of meaning representation can be reconstructed by BERT.

What emerges from these explorations is that BERT clearly captures fine-grained, local semantic similarities between tokens. Even with an entirely unsupervised application of BERT, discrete, coherent token groupings can be discerned that correspond relatively well with the sense categories proposed by linguists. Furthermore, embeddings of over also clearly encode information about conceptual domains, as concrete, spatial uses of prepositions are neatly distinguished from more abstract,metaphorical extensions (into the conceptual domain of time, or other non-spatial domains). However, there are no indications that BERT embeddings also encode information about the abstract image schema resemblances between tokens across those domains. These findings highlight the fact that such imagistic similarities may not be straightforwardly captured in contextualized. Such findings can provide an interesting basis for further experimental research (testing to what extent different operational models of meaning representation are complementary when assessed against elicited behavioral data), as well as a discussion on how we can bring about a "greater crossfertilization of theoretical and computational approaches" to the study of meaning (Boleda 2020: 213; Baroni & Lenci 2011).

Baroni, Marco, Georgiana Dinu & Germán Kruszewski. 2014. Don't count, predict! A systematic comparison of context-counting vs. context-predicting semantic vectors. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 238–247.
Boleda, Gemma. 2020. Distributional Semantics and Linguistic Theory. Annual Review of Linguistics 6(1). 213–234.
Devlin, Jacob, Ming-Wei Chang, Kenton Lee & Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL-HLT 2019. 4171–4186.
Tyler, Andrea & Vyvyan Evans. 2001. Reconsidering Prepositional Polysemy Networks: The Case of Over. Language 77(4). 724–765.