Data: dijous 25 de març del 2021

Hora: 12.00 h

Accés: en línia, amb Collaborate (enllaç: eu.bbcollab.com/guest/cbd918ec0de7403f94ba44b0a67b7a2d)

Resum:
The term ‘meaning’, as it is presently employed in Linguistics, is a polysemous concept, covering a broad range of operational definitions. Focussing on two of these definitions, meaning as ‘concept’ and meaning as ‘context’ (also known as ‘distributional semantics’), this paper explores to what extent these operational definitions lead to converging conclusions regarding the number and nature of distinct senses a polysemous form covers. More specifically, it investigates whether the sense network that emerges from the principled polysemy model of over as proposed by Tyler & Evans (2003; 2001) can be reconstructed by the neural (or predictive, see Baroni et al. 2014) language model BERT. The study assesses whether the contextual information encoded in BERT embeddings can be employed to successfully (i) recognize the abstract sense categories and (ii) replicate the relative distances between the senses of over proposed in the principled polysemy model.

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, as in (1) are neatly distinguished from more abstract, metaphorical extensions (into the conceptual domain of time, or other non-spatial domains), as in (2):

1. I noticed a painting hanging over the piano (COHA, 2006)
2. a. The war on witchcraft intensified over the next 200 years, sending millions of cats, not to mention humans, to their deaths. (COHA, 2001)
    b. But Mike had seemed okay with it, as if he was completely over Lindsey (COHA, 2009).

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 embeddings. 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 behavioural data), as well as a discussion on how we can bring about a "greater cross-fertilization of theoretical and computational approaches" to the study of meaning (Boleda 2020: 213).

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 Linguistics6(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.