Project 1

Corpus evidence for delineating constructions

In a usage-based construction grammar approach, “constructionhood” is a gradual phenomenon rather than a categorical choice. In addition, individual differences between users of a language further complicate decisions about the constructional status of patterns. Despite this, constructionist analyses and constructicon projects still require individual choices about what is considered a construction. On the form side, this involves decisions on which patterns are habitual or entrenched enough to warrant an analysis, where the borders of these patterns are, and which open, partially restricted, or fixed slots they have.

A popular family of quantitative methods in construction grammar is collostructional analysis, which consists of simple collexeme analysis (Stefanowitsch & Gries 2003), and its extensions distinctive collexeme analysis (Gries & Stefanowitsch 2004b) and co-varying collexeme analysis (Gries & Stefanowitsch 2004a). These three methods investigate various aspects related to the association between constructions, or constructional slots, and lexemes. Another matter closely related to contructional association is syntactic productivity (cf. Zeldes 2012 for an overview over different methods for measuring productivity).

While these methods shed light on the quantitative properties of a known pattern, there are also attempts to (semi-)automatically discover construction candidates in corpora. Most of these approaches focus on the form side and are based on frequency, association between constituents, and sometimes productivity (e.g. Forsberg et al. 2014, Dunn 2017). Some approaches also take the semantic side into account. Perek and Patten (2019) match COBUILD grammar patterns with FrameNet frames and then consider constructions to be abstractions over clustered FrameNet frames. Martí et al. 2019 base construction candidates on the relations between clusters of semantically related words.

Despite this variety of existing quantitative approaches in construction grammar, explicit quantitative criteria for deciding whether a given pattern is a construction, how far it extends, and at which level of schematicity a construction should be posited are rarely discussed.

The goal of this project is to collect, combine, motivate, and evaluate quantitatively measurable criteria of constructionhood to support linguistic and constructicographic decisions. This will involve finding solutions for combining different measures, comparing patterns of different lengths and levels of abstraction, and identifying and representing semantic constraints on open slots.

References:

Dunn, J. (2017). Computational learning of construction grammars. Language and cognition, 9(2), 254–292.

Forsberg, M., Johansson, R., Bäckström, L., Borin, L., Lyngfelt, B., Olofsson, J., & Prentice, J. (2014). From construction candidates to constructicon entries: An experiment using semi-automatic methods for identifying constructions in corpora. Constructions and Frames, 6. https://doi.org/10.1075/cf.6.1.07for

Gries, S. T., & Stefanowitsch, A. (2004a). Co-varying collexemes in the into-causative. Language, culture, and mind, 225–236.

Gries, S. T., & Stefanowitsch, A. (2004b). Extending collostructional analysis: A corpus-based perspective on ‘alternations’. International journal of corpus linguistics, 9(1), 97–129.

Martí, M., Delor, M., Kovatchev, V., & Salamó, M. (2019). Discover: Distributional approach based on syntactic dependencies for discovering constructions. Corpus Linguistics and Linguistic Theory, 17.https://doi.org/10.1515/cllt-2018-0028

Perek, F., & Patten, A. (2019). Towards an english constructicon using patterns and frames. International Journal of Corpus Linguistics, 24, 354–384. https://doi.org/10.1075/ijcl.00016.per

Stefanowitsch, A., & Gries, S. T. (2003). Collostructions: Investigating the interaction of words and constructions. International journal of corpus linguistics, 8(2), 209–243.

Zeldes, A. (2012). Productivity in argument selection: From morphology to syntax. De Gruyter Mouton. https://doi.org/doi:10.1515/9783110303919

This project is done by Malin Patel and supervised by Prof. Dr. Stephanie Evert and Prof. Dr. Thomas Herbst.