Nederlandse Taalkunde - Volume 25, Issue 1, 2020

Corpora are a useful and important source of evidence for linguistic research, but they are not the only kind of evidence, do not have any special status as evidence, and have their limitations. Recent very user-friendly applications such as GrETEL make it very easy to search in large and richly annotated corpora on the basis of an example sentence and without knowledge of a query language or the exact nature of the linguistic annotations. It is therefore very tempting to use these applications intensively. That is fine, but also dangerous in ways, because in many cases, in order to interpret the results correctly, the researcher must really be aware of the precise nature of the linguistic annotations and of the way in which the user-friendly interface generates a query on the basis of an example sentence. I will illustrate this with several examples. I also sketch some methods for avoiding or mitigating the dangers and argue that the applications should support these methods also in as user-friendly a manner as possible.

In this contribution, I discuss the use of automatic syntactic annotation in Dutch corpus research, using a case study of five-verb clusters. Large amounts of text can be annotated automatically, but the parser makes mistakes, while correct annotation is very important in linguistic research. How much of a problem is this, and how can we learn about the extent of these parsing mistakes? There are several approaches to evaluating the quality of automatic annotation for specific research questions. I demonstrate these approaches for the case study at hand, which will help us to make claims based on automatically annotated corpus data with greater confidence.

Belgian Dutch (BD) and Netherlandic Dutch (ND) are known to exhibit phonetic and lexical differences, but national variation in the syntax of Dutch has often been claimed to be quasi non-existent. This view is rooted in the fact that both laypersons and researchers are oblivious to national divergences in the grammar of Dutch (unless they are categorical and/or heavily mediatized), but also in the undisputed belief that BD and ND are different surface manifestations of ‘the same grammatical motor’. As a result, only a few syntactic phenomena have hitherto been shown to be sensitive to national constraints. In this paper we illustrate a computational bottom-up approach (pioneered in Bannard & Callison-Burch 2005) to cast the net as widely as possible. Building on statistical machine translation and a parallel corpus of Dutch translations of English subtitles, we identify plausible mappings between English n-grams and their Dutch translations. We do this in order to obtain paraphrases, i.e., stretches of interchangeable Dutch text that carry approximately the same meaning. In a first case study, we found corroborating evidence among the discovered paraphrases for many syntactic variables that have previously been attested in Dutch, including complementizer variation, existential er-variation, word order phenomena, and inflection variation. Crucially, we also discovered a number of alternations we had not anticipated as interesting variables. In order to detect national constraints on the newly found variables, we carried out a second experiment with a smaller corpus of Belgian and Netherlandic subtitles: the two variables we investigated in this light – deictic strength variation and subordination variation – did indeed manifest national sensitivity.

In this paper, we revisit earlier analyses of the distribution of er ‘there’ in adjunct-initial sentences to demonstrate the merits of computational upscaling in syntactic variation research. Contrary to previous studies, in which major semantic and pragmatic predictors (viz. adjunct type, adjunct concreteness, and verb specificity) had to be coded manually, the present study operationalizes these predictors on the basis of distributional analysis: instead of hand-coding for specific semantic classes, we determine the semantic class of the adjunct, verb, and subject automatically by clustering the lexemes in those slots on the basis of their ‘semantic passport’ (as established on the basis of their distributional behaviour in a reference corpus). These clusters are subsequently interpreted as proxies for semantic classes. In addition, the pragmatic factor ‘subject predictability’ is operationalized automatically on the basis of collocational attraction measures, as well as distributional similarity between the other slots and the subject. We demonstrate that the distribution of er can be modelled equally successfully with the automated approach as in manual annotation-based studies. Crucially, the new method replicates our earlier findings that the Netherlandic data are easier to model than the Belgian data, and that lexical collocations play a bigger role in the Netherlandic than in the Belgian data. On a methodological level, the proposed automatization opens up a window of opportunities. Most important is its scalability: it allows for a larger gamut of alternations that can be investigated in one study, and for much larger datasets to represent each alternation.

Standard Dutch, the key to integration in Flanders? About the challenges posed by non-standard Dutch to language learners in Flanders

Flanders strongly encourages newcomers to take language courses to learn the official language, on the assumption that certified competence in Dutch will enhance integration. However, in Dutch L2-education the focus lies on Standard Dutch, and this contrasts to everyday spoken encounters in Flanders, where Standard Dutch is rarely used. Despite the perceived importance of learning the official language of the host society, the attested contrast between policy and language reality has up to now not been given much attention. In this publication we go into the results of a linguistic-ethnographic study conducted with language learners living in East-Flanders. We show that non-standard Dutch used by L1-speakers of Dutch sometimes causes comprehension difficulties in their interactions with the language learners. Based on interactional and interview data, we reflect on the question if L2-education should pay more attention to non-standard Dutch in class.