CRC 1412/2: GeRMaN: German register marking by non-literal expressions (SP A01)

As part of Area A, we investigate the interdependence of register knowledge and grammar (Question QA),
in our case, register and pragmatics, as it emerges in the use and interpretation of non-literal expressions
(NLEs). We widen the perspective of Phase I from metaphor and metonymy to e.g. rhetorical questions,
irony, and indirect speech acts and include new oral text types. Exploring how semantic alternations like
choosing NLEs over alternative literal expressions are recruited for register purposes addresses question QAi.
We want to model the impact of register on the choice of NLE in the form of variables that can be integrated
into probabilistic models of pragmatic enrichment of meaning that include NLE (QAii and QAiii).
For the acquisition of the relevant data from linguistic corpora, we will explore new ways of NLE detection in corpora
through Deep Learning. This will contribute to QAi by safeguarding sufficient data for our exploration of
NLE use in different registers and constitute a modelling of register and its impact on grammar by itself
(QAiii). Our modelling is statistical in both senses of th e term employed in Section 1: We represent the
interdependence between register knowledge and NLE probabilistically, and render our results in the form
of variables for models of Bayesian pragmatics and in the form of Deep Learning models.
We will first annotate the full range of the NLEs in our corpus of six text types, which we compiled in Phase
I (Egg & Kordoni, 2022, accepted). At the same time, we will considerably extend the number of potentially
register-relevant NLEs by identifying a group of ‘conventionalised’ metaphors that can be used as additional
register markers and through a more fine-grained analysis of metaphorical multi-word expressions. Since
oral discourse emerged in the first phase as very interesting yet understudied phenomenon (with respect to
our approach to register), we will in a second step augment the corpus to include a wider variety of oral
dialogue, e.g., sales talks or oral examinations, falling back on transcripts of the FOLK corpus from the
Spoken German Database (Datenbank Gesprochenes Deutsch; Schmidt, 2014).
[...]