veranstalter: Prof. Dr. Christopher Kyle (University of Hawaii at Manoa)
ansprechpartner: Karin Madlener, David Schreiber
email: karin.madlener@unibas.ch
web:
institution: HPSL
language: Englisch
location institution: Basel
date_raw: 29. März 2017 14:00-16:30 Uhr
date_sort: 29.03.2017, 00:00:00

A scalable approach to the extraction of
constructions: Replicable (semi-) automatic techniques for analyzing linguistic
patterns

Constructions
are form-meaning pairs that exist at all levels of language (Goldberg, 1995). Constructions
in general, and verb argument constructions in particular have been of
considerable interest in L1 and L2 research from a usage-based perspective (e.g.,
Behrens et al., 2000; Ellis & Ferreira-Junior, 2009; Goldberg, et al.,
2004; Madlener, 2015). While some linguistic patterns (e.g., fixed multiword
sequences) are easily identifiable using common corpus tools/techniques, the
extraction of VACs has been much more labor intensive. For this reason, most
investigations of VAC use tend to explore relatively small datasets and/or a
limited number of constructions (e.g., Gries & Stefanowitsch, 2004; Römer,
O’Donnell, & Ellis, 2015).

An
important factor that has limited large scale VAC analysis has been the nature
and reliability of syntactic parsers. Over the past ten years however, there
has been an increased interest in dependency parsing (e.g., Briscoe, 2006;
Nivre, Hall, & Nilsson, 2006) and advances in syntactic parsing (e.g., Chen
& Manning, 2014) have significantly increased parsing accuracy. These two
factors have paved the way for large-scale, comprehensive analyses of VAC use
(Kyle, 2016). Such analyses have recently comprehensively cataloged VAC use in
large reference corpora such as the circa 450-million-word Corpus of
Contemporary American English (COCA; Davies, 2010). Automatic VAC analysis has
also been applied to learner corpora (Kyle, 2016; Kyle & Crossley, 2015;
2016), allowing for learner VAC use to be analyzed via reference corpus
frequency and strength of association.

In
this workshop, I will a) discuss approaches to the automatic extraction of
linguistic patterns, b) report on two projects that investigate verb argument
construction (VAC) use longitudinally and across proficiency levels, and c)
provide a hands-on demonstration of the use of the Tool for the Automatic Analysis
of Syntactic Sophistication and Complexity (TAASSC).

Prof. Dr. Kristopher Kyle ist aktuell Assistenzprofessor am Department of Second Language Studies der University of Hawaii at Manoa, USA. Er promovierte zuvor an der Georgia State University in Atlanta, USA. Prof. Kyle beschäftigt sich mit korpus- und computerlinguistischen Fragestellungen in verschiedenen Bereichen des Erst- und Zweisprachgebrauchs und -erwerbs, z.B. mit syntaktischer Komplexität, lexikalischer Variabilität, textueller Kohäsion und (semi-) automatischen Bewertungsmöglichkeiten für Texte. Er hat selbst einschlägige korpuslinguistische Tools entwickelt.

Weitere Informationen zum Dozenten finden Sie hier und hier.