Relative clauses as a benchmark for Minimalist parsing

Abstract

Minimalist grammars have been used recently in a series of papers to explain well-known contrasts in human sentence processing in terms of subtle structural differences. These proposals combine a top-down parser with complexity metrics that relate parsing difficulty to memory usage. So far, though, there has been no large-scale exploration of the space of viable metrics. Building on this earlier work, we compare the ability of 1600 metrics to derive several processing effects observed with relative clauses, many of which have been proven difficult to unify. We show that among those 1600 candidates, a few metrics (and only a few) can provide a unified account of all these contrasts. This is a welcome result for two reasons: First, it provides a novel account of extensively studied psycholinguistic data. Second, it significantly limits the number of viable metrics that may be applied to other phenomena, thus reducing theoretical indeterminacy.

Testing a Minimalist Grammar Parser on

“…Extending previous work on MG parsing (Kobele et al, 2013;Graf and Marcinek, 2014;Gerth, 2015), Graf et al (2017) distinguish three cognitive notions of memory usage: I) how long a node is kept in memory (tenure); II) how many nodes must be kept in memory (payload); or III) how many bits a node consumes in memory (size). Tenure and payload for each node n in the tree can be easily computed via the node annotation scheme of Kobele et al: a node's tenure is equal to the difference between its index and its outdex; the payload of a derivation tree is computed as the number of nodes with a tenure strictly greater than 2.…”

“…In order to do so, these notions of memory have been used to define a vast set of complexity metrics measuring processing difficulty over a full derivation tree. In this paper, we look at Italian relative clause asymmetries using the full set of 1600 metrics as defined in Graf et al (2017). However, in what follows we only give a general intuition of how such metrics can be defined, and we refer the reader to Graf et al for the detailed formal definitions.…”

“…, M n , similar to constraint ranking in Optimality Theory (Prince and Smolensky, 2008): a lower ranked metric matters only if all higher ranked metric have failed to pick out a unique winner (e.g., if two constructions result in a tie over MAXT). This suggestion is fully explored in Graf et al (2017), which show that when complexity metrics are allowed to be ranked in such a way the total number of possible metrics quickly reaches an astronomical size. However, surveying the variety of previously modeled phenomena, the authors also suggest that the number of metrics truly needed to account for human processing contrasts can be reduced to a small number of core metrics (particularly, they point toward a combination of MAXT and SUMS), an hypothesis that seems supported by recent work on several different constructions (Liu, 2018;Lee, 2018).…”

“…Recent studies have shown that a top-down parser for Minimalist grammars (MGs;Stabler, 1996Stabler, , 2013) can be combined with complexity metrics to relate parsing behavior to memory usage, and successfully used to model sentence processing preferences across a variety of phenomena crosslinguistically (Kobele et al, 2013;Gerth, 2015;Graf et al, 2017). This kind of work follows a line of research on syntactic processing that sees computational models provide a transparent, interpretable linking theory between syntactic assumptions and processing behavior (Joshi, 1990;Rambow and Joshi, 1994;Hale, 2011).…”

“…Here, I adopt Kobele et al's (2013) implementation of Stabler's (2013) top-down traversal algorithm, coupled with the set of complexity metrics defined by Graf et al (2017). We test the MG parser's performance on the processing asymmetries reported for Italian relative clauses, which have been object of extensive study in the psycholinguistic literature.…”