Volume 16, Issue 2 -

The aim of this study is to explore the possibility of identifying speaker stance in discourse, provide an analytical resource for it and an evaluation of the level of agreement across speakers. We also explore to what extent language users agree about what kind of stances are expressed in natural language use or whether their interpretations diverge. In order to perform this task, a comprehensive cognitive-functional framework of ten stance categories was developed based on previous work on speaker stance in the literature. A corpus of opinionated texts was compiled, the Brexit Blog Corpus (BBC). An analytical protocol and interface (Active Learning and Visual Analytics) for the annotations was set up and the data were independently annotated by two annotators. The annotation procedure, the annotation agreements and the co-occurrence of more than one stance in the utterances are described and discussed. The careful, analytical annotation process has returned satisfactory inter- and intra-annotation agreement scores, resulting in a gold standard corpus, the final version of the BBC.

This paper explores the proposed benefits of Δ P (delta P ) as a measure of collocation strength. Its focus is on contrasting Δ P with other, more commonly used, association measures, particularly transitional probabilities, but also mutual information and Lexical Gravity G . To this end, first the strong correlation between Δ P and transitional probability is illustrated with the help of two exemplary corpora. This is followed by an analysis of hesitation placement in spontaneous spoken English, based on the assumption that hesitations will not be placed within strong collocations. Results show that, despite their strong similarity, in some contexts Δ P is more predictive of hesitation placement than transitional probability. Yet neither Δ P nor any of the other association measures emerges as the universally best predictor. On the basis of these results, it is suggested that studies should always rely on several association measures.

This study present a corpus-based comparison of two aspectual-sematic classification models proposed in theoretical literature (unidimensional vs. bidimensional) by applying them to a set of nominal and verbal gerunds from the Modern English period. It (i) summarises the differences between unidimensional and bidimensional classification models and (ii) the potential problems associated with them. Despite the difficulties of studying semantic aspect in Present-day as well as historical data, this study will argue that, (iii) at least for deverbal nominalization patterns, it is possible to take a bidimensional approach and maintain a clear distinction between, on the one hand, aspect features of the nominalized situation (stativity/dynamicity, durativity/punctuality, and telicity/atelicity), and temporal boundedness of that situation. The question of which semantic classification model to use, then, is not so much one of which one is practically feasible in a corpus analysis, but rather which one is best suited to describe the attested variation. In order to determine the best model (in terms of parsimony and descriptive accuracy), (iv) the models were compared by means of ‘akaike weights’. To describe the variation between nominal and verbal gerunds in Early and Late Modern English, the bidimensional model outperformed the unidimensional one, showing that (v) the aspectual-semantic distinctions between Modern English nominal and verbal gerunds are a matter of both aspect and temporal boundedness .

This paper draws on data from learner and native-speaker corpora as well as psycholinguistic data to gain insights into second language speaker knowledge of English verb-argument constructions (VACs). For each of 34 VACs, L1 German and L1 Spanish advanced English learners’ and English native speakers’ dominant verb–VAC associations are examined based on data retrieved from the International Corpus of Learner English (ICLE), the Louvain International Database of Spoken English Interlanguage (LINDSEI), their respective Native Speaker (NS) reference corpora, and data collected in verbal fluency tasks in which participants complete VAC frames, such as, ‘she _______ with the…’ with verbs that come to mind. We compare findings from the different data sets and consider the strengths and limitations of each in relation to questions in usage-based second language acquisition and Construction Grammar.

It is widely assumed that the frequency adverb always combined with the progressive aspect is typically used in negative evaluations expressing irritation, i.e., complaints. Adopting a cognitive-functional approach, I test this claim across six genres of Present Day English. Always progressives were coded according to their functions: Describe (neutral), Complain (negative), Lament (negative), or Praise (positive). Neutral, rather than negative, functions predominated in all genres, although negative functions outnumbered positive functions. I relate the former finding to the propensity of always progressives to act similarly to the simple aspect and the latter to a cognitive phenomenon called the negativity bias.

In the context of constructional alternatives, we may assume that speakers’ choice between alternative forms is influenced by a multitude of factors. At the moment, multivariate statistical classification modelling seems to be the best tool available to capture this knowledge quantitatively. There is a vast array of techniques available. In this paper, two distinct modelling techniques are applied – logistic regression and naïve discriminative learning – to predict the choice between two constructional alternatives in written Estonian. One of the central questions in statistical modelling concerns the evaluation of model fit. It is proposed that for linguistic analysis, the performance of alternative corpus-based models can be evaluated by, first, pitting them against each other and second, pitting them against experimental data. Previous work on modelling constructional and lexical choice has focused on one of the two aspects. The present paper takes this line of analysis further by combining the two approaches.

This paper presents token-based semantic vector spaces as a tool that can be applied in corpus-linguistic analyses such as word sense comparisons, comparisons of synonymous lexical items, and matching of concordance lines with a given text. We demonstrate how token-based semantic vector spaces are created, and we illustrate the kinds of result that can be obtained with this approach. Our main argument is that token-based semantic vector spaces are not only useful for practical corpus-linguistic applications but also for the investigation of theory-driven questions. We illustrate this point with a discussion of the asymmetric priming hypothesis (Jäger and Rosenbach 2008). The asymmetric priming hypothesis, which states that grammaticalizing constructions will be primed by their lexical sources but not vice versa, makes a number of empirically testable predictions. We operationalize and test these predictions, concluding that token-based semantic vector spaces yield conclusions that are relevant for linguistic theory-building.