Search

This article focuses on the practical problems that go along with the sequencing of older non-standardised spelling forms into consonant and vowel graphemes. The issue of segmental parsing is important in research on the development of Dutch spelling, mostly drawing from thirteenth and fourteenth century charter spellings. The research aims are introduced in part 1, and then the problems of defining graphs and then syllables are discussed in part 2. These problems stem from the theoretical graphemic level, but have a very direct impact on the practical level, as they hinder the automatic parsing of tokens in the corpus, as discussed in part 3. The goal of this article is to provide partial solutions in parsing of non-standardised language data for graphemic research.

Shortening is a common type of word-formation in many languages. Crystal (2008) distinguishes two kinds of abbreviation: initialisms and acronyms. Article use in English is variable with both acronyms and initialisms used as proper names (e.g. (the) UKIP, at the UN vs. at MIT). The question is whether variability is largely dependent on the semantics of the underlying full form (i.e. whether this is derived from a proper name or common noun) or whether the two types of abbreviation show different behaviour with respect to variable article use. This paper uses data from CoStEP, a new, word-aligned version of EuroParl, and a data-driven approach to investigate variable article use with abbreviations and their full forms uttered by English native speakers and compares the findings to data from parallel German and Italian corpora. The results show higher article variability in English and a marked preference for and near categorical article use in German and Italian. Furthermore, our evidence confirms that acronyms tend towards the proper name end of the cline, while initialisms behave syntactically more like common nouns.

This study implements an automated parser-based approach to the investigation of light verb constructions. The database consisting of ICE-GB and ICE-IRE is used to obtain qualitative and quantitative results on the use of light verb structures. The study explains and evaluates the steps employed to optimize parser output in detecting open lists of light verb constructions. It discusses the qualitative usage differences of these structures in the data between the two varieties and finds that ICE-GB favours fewer high frequency light verbs while ICE-IRE contains more diverse lower frequency light verbs and more passives. Overall, counts of light verb constructions are considerably higher than previously assumed. The projected counts suggest that attestations of light verb constructions will increase considerably if the search is not restricted to certain high-frequency light verbs as is typically done in studies employing manual or semi-automatic approaches to data collection.