Dutch Corpus Research Articles

Interjections and the Language Functions Debate

Five views of the function of interjections, developed in the first half of the 20 th century by the psychologist-linguist Buhler, the linguists Gardiner, and Jakobson, and the psychologists Revesz and Duijker, are discussed. All five scholars reject the earlier psychologism that reinforced the traditional emotion-expressive view of interjections; all of them argue for a listener-directed, communicative view of language in general, and all include a specific appeal-to-the-listener-function in their model of language functions. My original hypothesis was therefore that these scholars would reject the one-sided traditional view that interjections mainly express the speaker’s emotions, acknowledging instead that the central function of most interjections is to make some appeal to the listener (a view supported by recent investigation of a corpus of spoken Dutch, which shows that only 7% fulfils a purely expressive function). As it turns out, however, all five scholars support the traditional view and attributed an expressive function to interjections. In this paper I try to explain this unexpected result.

Multilevel modeling of between-speaker and within-speaker variation in spontaneous speech tempo

Speech tempo (articulation rate) varies both between and within speakers. The present study investigates several factors affecting tempo in a corpus of spoken Dutch, consisting of interviews with 160 high-school teachers. Speech tempo was observed for each phrase separately, and analyzed by means of multilevel modeling of the speaker's sex, age, country, and dialect region (between speakers) and length, sequential position of phrase, and autocorrelated tempo (within speakers). Results show that speech tempo in this corpus depends mainly on phrase length, due to anticipatory shortening, and on the speaker's country, with different speaking styles in The Netherlands (faster, less varied) and in Flanders (slower, more varied). Additional analyses showed that phrase length itself is shorter in The Netherlands than in Flanders, and decreases with speaker's age. Older speakers tend to vary their phrase length more (within speakers), perhaps due to their accumulated verbal proficiency.

Automatic phonetic transcription of large speech corpora

Most large speech corpora are delivered with a lexicon that contains a canonical transcription of every word in the orthographic transcription. Such a lexicon can be used for generating a hypothetical ‘canonical’ phonetic transcription from the orthography. In addition, time and money permitting, some speech corpora are provided with a manually verified broad phonetic transcription of at least part of the material. Since the manual verification of phonetic transcriptions is time-consuming and expensive, we investigated whether existing automatic transcription procedures and combinations of such procedures can offer a quick and cheap alternative for the generation of phonetic transcriptions like the manually verified transcriptions delivered with large speech corpora. In our study, we used 10 automatic transcription procedures to generate a broad phonetic transcription of well-prepared speech (read-aloud texts) and spontaneous speech (telephone dialogues) from the Spoken Dutch Corpus. The performance was assessed in terms of the number and the nature of the discrepancies between the emerging phonetic transcriptions and the corresponding manually verified phonetic transcriptions delivered with the Spoken Dutch Corpus. Some of the resulting automatic transcriptions appeared to be comparable to the manually verified transcriptions.

On the Utility of Syllable-Based Acoustic Models for Pronunciation Variation Modelling

Recent research on the TIMIT corpus suggests that longer-length acoustic models are more appropriate for pronunciation variation modelling than the context-dependent phones that conventional automatic speech recognisers use. However, the impressive speech recognition results obtained with longer-length models on TIMIT remain to be reproduced on other corpora. To understand the conditions in which longer-length acoustic models result in considerable improvements in recognition performance, we carry out recognition experiments on both TIMIT and the Spoken Dutch Corpus and analyse the differences between the two sets of results. We establish that the details of the procedure used for initialising the longer-length models have a substantial effect on the speech recognition results. When initialised appropriately, longer-length acoustic models that borrow their topology from a sequence of triphones cannot capture the pronunciation variation phenomena that hinder recognition performance the most.

The role of morphological family size in word recognition

This paper proposes an approach for studying the structure and development of the mental lexicon based on morphological family size. For adults, the number of morphologically related words has been shown to facilitate word recognition (Schreuder, & Baayen, 1997). This effect is assumed to be caused by connections in the mental lexicon between morphologically related words. Because children still have to acquire word representations and their interconnectedness, family size effects in children are considered to be smaller. So far, however, developmental data on morphological family size effects are generally lacking. As a first step, we wanted to find out to what extent word frequency and morphological family size effects will appear in adult lexical decision data while using adult versus child language corpora as a frame of reference. Two findings were obtained. First, the materials specifically constructed for children evoked both frequency and family size effects. Second, frequency counts based on a Dutch corpus of child language (Schrooten & Vermeer, 1994) showed reliable frequency effects.

Belgian Dutch versus Netherlandic Dutch: New patterns of divergence? On pronouns of address and diminutives

The linguistic climate in northern Belgium (Flanders) has been changing in recent years. A new corpus of spoken Dutch meets the need for data reflecting actual and present-day language use in this part of the Dutch language area. The ‘Spoken Dutch Corpus’ allows us to uncover and analyse the present state of colloquial Belgian Dutch and the changes which mark this condition. This paper discusses the realization of two morphosyntactic variables, the variants of which are markers of Belgian Dutch versus Netherlandic Dutch. In spite of more than half a century of official language policy promoting convergence with northern Netherlandic Dutch, the results reveal a growing divergence from the northern norm: the younger generations show a greater preference for the endogenous, Brabantic, variants than the older generations. The northern Dutch variants have not been integrated in colloquial Belgian Dutch. Apparently they are still considered to be ‘Hollandic import’. 1

An elitist approach to automatic articulatory-acoustic feature classification for phonetic characterization of spoken language

A novel framework for automatic articulatory-acoustic feature extraction has been developed for enhancing the accuracy of place- and manner-of-articulation classification in spoken language. The “elitist” approach provides a principled means of selecting frames for which multi-layer perceptron, neural-network classifiers are highly confident. Using this method it is possible to achieve a frame-level accuracy of 93% on “elitist” frames for manner classification on a corpus of American English sentences passed through a telephone network (NTIMIT). Place-of-articulation information is extracted for each manner class independently, resulting in an appreciable gain in place-feature classification relative to performance for a manner-independent system. A comparable enhancement in classification performance for the elitist approach is evidenced when applied to a Dutch corpus of quasi-spontaneous telephone interactions (VIOS). The elitist framework provides a potential means of automatically annotating a corpus at the phonetic level without recourse to a word-level transcript and could thus be of utility for developing training materials for automatic speech recognition and speech synthesis applications, as well as aid the empirical study of spoken language.

Multiword expressions in spoken language: An exploratory study on pronunciation variation

The study presented in this paper was aimed at exploring the possibilities of modelling specific pronunciation characteristics of multiword expressions (MWEs) for both automatic speech recognition (ASR) and automatic phonetic transcription (APT). For this purpose, we first drew up an inventory of frequently found N-grams extracted from orthographic transcriptions of spontaneous speech contained in a large corpus of spoken Dutch. These N-grams were filtered and subsequently assigned to linguistic categories. For a small selection of these N-grams we examined the phonetic transcriptions contained in the corpus. We found that the pronunciation of these N-grams differed to a large extent from the canonical form. In order to determine whether this is a general characteristic of spontaneous speech or rather the effect of the specific status of these N-grams, we analysed the pronunciations of the individual words composing the N-grams in two context conditions: (1) in the N-gram context and (2) in any other context. We found that words in N-grams do indeed have peculiar pronunciation patterns. This seems to suggest that the N-grams investigated may be considered as MWEs that should be treated as lexical entries in the pronunciation lexicons used in ASR and APT, with their own specific pronunciation variants.

A Controlled-corpus Experiment in Authorship Identification by Cross-entropy

This paper describes an authorship, and more generally document classification, experiment on a preexisting Dutch corpus of university writings. By measuring linguistic distances using a cross-entropy technique, a technique sensitive not only to the distributions of language features, but also to their relative interse- quencing, classification judgments can be made with great sensitivity, significance, confidence, and accuracy. In particular, despite the designed diculty of the Dutch corpus used, the technique was still able to reliably detect not only authorship, but also subtle features of register, topic, and even the educational attainments of the author. We present evidence suggesting that this technique outperforms more well-known techniques such as function word principal components analysis or linear discriminant analysis, as well as suggest ways in which performance can be improved.

Lexical frequency and voice assimilation in complex words in Dutch

Words with higher token frequencies tend to have more reduced acoustic realizations than lower frequency words (e.g., Hay, 2000; Bybee, 2001; Jurafsky et al., 2001). This study documents frequency effects for regressive voice assimilation (obstruents are voiced before voiced plosives) in Dutch morphologically complex words in the subcorpus of read-aloud novels in the corpus of spoken Dutch (Oostdijk et al., 2002). As expected, the initial obstruent of the cluster tends to be absent more often as lexical frequency increases. More importantly, as frequency increases, the duration of vocal-fold vibration in the cluster decreases, and the duration of the bursts in the cluster increases, after partialing out cluster duration. This suggests that there is less voicing for higher-frequency words. In fact, phonetic transcriptions show regressive voice assimilation for only half of the words and progressive voice assimilation for one third. Interestingly, the progressive voice assimilation observed for higher-frequency complex words renders these complex words more similar to monomorphemic words: Dutch monomorphemic words typically contain voiceless obstruent clusters (Zonneveld, 1983). Such high-frequency complex words may therefore be less easily parsed into their constituent morphemes (cf. Hay, 2000), favoring whole word lexical access (Bertram et al., 2000).

The effect of gender on the choice of speech style

Abstract The analysis of a number of linguistic variables reflecting formality in the advanced oral French interlanguage of 27 Dutch-speaking students showed significant differences between male and female speakers. Female speakers were found to opt for a much more implicit and deictical speech style, especially in an informal situation. As the situation became more formal, the differences between male and female speakers weakened. The same variation was found in a corpus of native Dutch.

Productivity in context: a case study of a Dutch suffix

The Dutch suffix -heid, like -ness in English, forms abstract nouns from adjectives. In this paper, we explore the hypothesis that -heid gives rise to two kinds of abstract nouns: on the one hand, nouns referring to concepts, and on the other hand, nouns referring to states of affairs. An examination of a corpus of newspaper Dutch reveals that the referential function of -heid is typical for the lowest-frequency words, most of which are neologisms. Conversely, its conceptual function is found predominantly among the highest-frequency words. Detailed investigations of the use of these two sorts of words in context showed that the high-frequency words tend to be less well anchored in their context than the low-frequency words, and that they pattern more as independent units. Our data argue against the view that productive word formation goes hand in hand with the absence of any storage of full forms in the mental lexicon. Instead, we claim that high-frequency formations with the productive suffix -heid are available in the mental lexicon, whereas low-frequency words and neologisms are produced and understood by rule

Politeness markers in French and Dutch requests

This paper describes a difference in social norms which results in a differing distribution of politeness markers in the Dutch and French speech communities. The use of politeness strategies is compared on the basis of a corpus of parallel Dutch and French requests. The differences in use can be explained by a different conception of the need for repair strategies.

Crosslinguistic Influence and the Expression of Hypothetical Meaning

Although Dutch learners of English generally achieve high levels of proficiency, even advanced students produce errors like the following: English Dutch *Ifshe would do that, I would leave her (Als zij dat zou doen, zou ik hij haar weg-gaan) instead of: If she did that, I would leave her (Als zij dat deed, zou ik bij haar weggaan) They tend to insert would in the ¿/^clause of hypothetical sentences. This type of error is a characteristic feature of Dutch English and in many cases it has become fossilized. The following questions present themselves: Why is it exactly this error which is so persistent while other aspects of grammar are acquired much more easily? To what extent does the mother tongue influence the fossilization process and to what extent do general principles like markedness and semantic disambiguation play a role? Dutch corpus data show that although the structural equivalent of both the correct and the incorrect structures above are perfectly acceptable in Dutch, one form is often highlyy preferred over the other, depending on factors like time frame (future/present/past), clause order and nature of the verb (strong/weak). On the basis of these facts, I will provide a thorough description of hypothetical conditionals, which serves as a starting point for my research. In order to gain more insight into the mother tongue in the fossilization process, a written test was administered to Dutch learners of English with different levels of proficiency, first in Dutch and a couple of weeks later in English. The learners had to fill in the verb forms in hypothetical conditional sentences embedded in a short situation description. The results showed that the distribution of the different forms in Dutch was to a large extent mirrored in English; i.e. where in Dutch the would equivalent was used, the number of if...would errors increased in English and where a past tense in the if-clause was favoured η Dutch, fewer mistakes were made in English. Apparently crosslinguistic influence does play a role here and the fossilization of if ..would seems to be limited only to a subset of conditional types.

"Een beetje babbelen onder elkaar"

This contribution opens wich some general information on the registration ofspeech and the use of corpora. Then the focus is on spontaneous speech: howis "spontaneous speech" defined, why is it so important to study this type ofspeech and which problems can arise when scudying it? Next, we focus oncorpora of spontaneous speech: (a) corpora created before the realization ofthe Spoken Ducch Corpus ("Corpus Gesproken Nederlands" or CGN), (b)the CGN itself and (c) a subcorpus of the CGN, the so-called "teachers'corpus".Finally we discuss some problems which arose in two studies, based onspontaneous speech from the "teachers' corpus": a study on vowel reductionand a study on speech rate. Keywords: Dutch, (corpora of) spontaneousspeech, Spoken Dutch Corpus, language variation.

Chattaal, tienertaal en taalverandering

This paper analyses the realisation of two morpho-syntactic variables, thevariants of which are markers of Belgian Dutch versus Netherlandic Dutch.The database consists of a corpus of chat language produced by Flemishteenagers. Both a quantitative analysis of the relative frequency of the variantsin the 'speech' of the Flemish teenagers and a qualitative analysis of theirfunctional-pragmatic distribution reveal a strong dominance of the endogenousBrabantic Dutch variants. Data from the Spoken Dutch Corpus fromthe spontaneous speech of young Flemish adults confirm this tendency. Inspite of more than half a century of official language policy promoting convergencewith northern Netherlandic Dutch, the northern standard Dutchvariants are hardly integrated in the peergroup language of Flemish teenagersand young adults, although they do belong to their linguistic repertoire.