Publications

Abstract

The event-related potential (ERP) component named phonological mismatch negativity (PMN) arises when listeners hear an unexpected word form in a spoken sentence [1]. The PMN is thought to reflect the mismatch between expected and perceived auditory speech input. In this paper, we use the PMN to test a central premise in the predictive coding framework [2], namely that the mismatch between prior expectations and sensory input is an important mechanism of perception. We test this with natural speech materials containing approximately 50,000 word tokens. The corresponding EEG-signal was recorded while participants (n = 48) listened to these materials. Following [3], we quantify the mismatch with two word probability distributions (WPD): a WPD based on preceding context, and a WPD that is additionally updated based on the incoming audio of the current word. We use the between-WPD cross entropy for each word in the utterances and show that a higher cross entropy correlates with a more negative PMN. Our results show that listeners anticipate auditory input while processing each word in naturalistic speech. Moreover, complementing previous research, we show that predictive language processing occurs across the whole probability spectrum.

Abstract

The mismatch between top-down predicted and bottom-up perceptual input is an important mechanism of perception according to the predictive coding framework (Friston, [1]). In this paper we develop and validate a new information-theoretic measure that quantifies the mismatch between expected and observed auditory input during speech processing. We argue that such a mismatch measure is useful for the study of speech processing. To compute the mismatch measure, we use naturalistic speech materials containing approximately 50,000 word tokens. For each word token we first estimate the prior word probability distribution with the aid of statistical language modelling, and next use automatic speech recognition to update this word probability distribution based on the unfolding speech signal. We validate the mismatch measure with multiple analyses, and show that the auditory-based update improves the probability of the correct word and lowers the uncertainty of the word probability distribution. Based on these results, we argue that it is possible to explicitly estimate the mismatch between predicted and perceived speech input with the cross entropy between word expectations computed before and after an auditory update.

Abstract

Previous research has demonstrated that language use can vary depending on the context of situation. The present paper extends this finding by comparing word predictability differences between 14 speech registers ranging from highly informal conversations to read-aloud books. We trained 14 statistical language models to compute register-specific word predictability and trained a register classifier on the perplexity score vector of the language models. The classifier distinguishes perfectly between samples from all speech registers and this result generalizes to unseen materials. We show that differences in vocabulary and sentence length cannot explain the speech register classifier’s performance. The combined results show that speech registers differ in word predictability.

Abstract

Previous studies have shown that speakers align their speech to each other at multiple linguistic levels. This study investigates whether alignment is mostly the result of priming from the immediately preceding
speech materials, focussing on pitch and articulation rate (AR). Native Dutch speakers completed sentences, first by themselves (pre-test), then in alternation with Confederate 1 (Round 1), with Confederate 2 (Round 2), with Confederate 1 again
(Round 3), and lastly by themselves again (post-test). Results indicate that participants aligned to the confederates and that this alignment lasted during the post-test. The confederates’ directly preceding sentences were not good predictors for the participants’ pitch and AR. Overall, the results indicate that alignment is more of a global effect than a local priming effect.

Abstract

This paper shows that the dictation task, a well-
known testing instrument in language education, has
untapped potential as a research tool for studying
speech perception. We describe how transcriptions
can be scored on measures of lexical, orthographic,
phonological, and semantic similarity to target
phrases to provide comprehensive information about
accuracy at different processing levels. The former
three measures are automatically extractable,
increasing objectivity, and the middle two are
gradient, providing finer-grained information than
traditionally used. We evaluate the measures in an
English dictation task featuring phonetically reduced
continuous speech. Whereas the lexical and
orthographic measures emphasize listeners’ word
identification difficulties, the phonological measure
demonstrates that listeners can often still recover
phonological features, and the semantic measure
captures their ability to get the gist of the utterances.
Correlational analyses and a discussion of practical
and theoretical considerations show that combining
multiple measures improves the dictation task’s
utility as a research tool.

Abstract

Lexically guided perceptual learning has traditionally been studied with ambiguous consonant sounds to which native listeners are exposed in a purely receptive listening context. To extend previous research, we investigate whether lexically guided learning applies to a vowel shift encountered by non-native listeners in an interactive dialogue. Dutch participants played a two-player game in English in either a control condition, which contained no evidence for a vowel shift, or a lexically constraining condition, in which onscreen lexical information required them to re-interpret their interlocutor’s /ɪ/ pronunciations as representing /ε/. A phonetic categorization pre-test and post-test were used to assess whether the game shifted listeners’ phonemic boundaries such that more of the /ε/-/ɪ/ continuum came to be perceived as /ε/. Both listener groups showed an overall post-test shift toward /ɪ/, suggesting that vowel perception may be sensitive to directional biases related to properties of the speaker’s vowel space. Importantly, listeners in the lexically constraining condition made relatively more post-test /ε/ responses than the control group, thereby exhibiting an effect of lexically guided adaptation. The results thus demonstrate that non-native listeners can adjust their phonemic boundaries on the basis of lexical information to accommodate a vowel shift learned in interactive conversation.

Abstract

An important dimension of linguistic variation is formality. This study investigates the role of social distance between interlocutors. Twenty-five native Dutch speakers retold eight short films to confederates, who acted either formally or informally. Speakers were familiarized with the informal confederates, whereas the formal confederates remained strangers. Results show that the two types of interlocutors elicited different versions of the same stories. Formal interlocutors (large social distance) elicited lower articulation rates, and more nouns and prepositions, both indicators of explicit information. Speakers addressing interlocutors to whom social distance was small, however, provided more explicit information with an involved character (i.e. adjectives with subjective meanings). They also used the word and more often as a gap filler or as a way to keep the floor. Furthermore, they were more likely to laugh and to use more interjections, first-person pronouns and direct speech, which are all indicators of involvement, empathy and subjectivity.

Abstract

Lombard speech, speech produced in noise, is acoustically different from speech produced in quiet (plain speech) in several ways, including having a higher and wider F0 range (pitch). Extensive research on native Lombard speech does not consider that non-natives experience a higher cognitive load while producing
speech and that the native language may influence the non-native speech. We investigated pitch range in plain and Lombard speech in native and non-natives.
Dutch and American-English speakers read contrastive question-answer pairs in quiet and in noise in English, while the Dutch also read Dutch sentence pairs. We found that Lombard speech is characterized by a wider pitch range than plain speech, for all speakers (native English, non-native English, and native Dutch).
This shows that non-natives also widen their pitch range in Lombard speech. In sentences with early-focus, we see the same increase in pitch range when going from plain to Lombard speech in native and non-native English, but a smaller increase in native Dutch. In sentences with late-focus, we see the biggest increase for the native English, followed by non-native English and then native Dutch. Together these results indicate an effect of the native language on non-native Lombard speech.

Abstract

Lombard speech, speech produced in noise, is
typically produced with a higher fundamental
frequency (F0, pitch) compared to speech in quiet. This paper examined the potential differences in native and non-native Lombard speech by analyzing median pitch in sentences with early- or late-focus produced in quiet and noise. We found an increase in pitch in late-focus sentences in noise for Dutch speakers in both English and Dutch, and for American-English speakers in English. These results
show that non-native speakers produce Lombard speech, despite their higher cognitive load. For the early-focus sentences, we found a difference between the Dutch and the American-English speakers. Whereas the Dutch showed an increased F0 in noise
in English and Dutch, the American-English speakers did not in English. Together, these results suggest that some acoustic characteristics of Lombard speech, such as pitch, may be language-specific, potentially
resulting in the native language influencing the non-native Lombard speech.

Abstract

Humans learn language by interaction with their environment and listening to other humans. It should also be possible for computational models to learn language directly from speech but so far most approaches require text. We improve on existing neural network approaches to create visually grounded embeddings for spoken utterances. Using a combination of a multi-layer GRU, importance sampling, cyclic learning rates, ensembling and vectorial self-attention our results show a remarkable increase in image-caption retrieval performance over previous work. Furthermore, we investigate which layers in the model learn to recognise words in the input. We find that deeper network layers are better at encoding word presence, although the final layer has slightly lower performance. This shows that our visually grounded sentence encoder learns to recognise words from the input even though it is not explicitly trained for word recognition.

Abstract

In order to study the cognitive processes underlying speech comprehension, neuro-physiological measures (e.g., EEG and MEG), or behavioural measures (e.g., reaction times and response accuracy) can be applied. Compared to behavioural measures, EEG signals can provide a more fine-grained and complementary view of the processes that take place during the unfolding of an auditory stimulus.

EEG signals are often analysed after having chosen specific time windows, which are usually based on the temporal structure of ERP components expected to be sensitive to the experimental manipulation. However, as the timing of ERP components may vary between experiments, trials, and participants, such a-priori defined analysis time windows may significantly hamper the exploratory power of the analysis of components of interest. In this paper, we explore a wide-window analysis method applied to EEG signals collected in an auditory repetition priming experiment.

This approach is based on a bank of temporal filters arranged along the time axis in combination with linear mixed effects modelling. Crucially, it permits a temporal decomposition of effects in a single comprehensive statistical model which captures the entire EEG trace.

Abstract

Many psycholinguistic models of speech sequence planning make claims about the onset and offset times of planning units, such as words, syllables, and phonemes. These predictions typically go untested, however, since psycholinguists have assumed that the temporal dynamics of the speech signal is a poor index of the temporal dynamics of the underlying speech planning process. This article argues that this problem is tractable, and presents and validates two simple metrics that derive planning unit onset and offset times from the acoustic signal and articulatographic data.

Abstract

Acoustic reduction refers to the frequent phenomenon in conversational speech that words are produced with fewer or lenited segments compared to their citation forms. The few published studies on the production and comprehension of acoustic reduction have important implications for the debate on the relevance of abstractions and exemplars in speech processing. This article discusses these implications. It first briefly introduces the key assumptions of simple abstractionist and simple exemplar-based models. It then discusses the literature on acoustic reduction and draws the conclusion that both types of models need to be extended to explain all findings. The ultimate model should allow for the storage of different pronunciation variants, but also reserve an important role for phonetic implementation. Furthermore, the recognition of a highly reduced pronunciation variant requires top down information and leads to activation of the corresponding unreduced variant, the variant that reaches listeners’ consciousness. These findings are best accounted for in hybrids models, assuming both abstract representations and exemplars. None of the hybrid models formulated so far can account for all data on reduced speech and we need further research for obtaining detailed insight into how speakers produce and listeners comprehend reduced speech.

Abstract

This article introduces a new speech corpus, the Nijmegen Corpus of Casual Czech (NCCCz), which contains more than 30 hours of high-quality recordings of casual conversations in Common Czech, among ten groups of three male and ten groups of three female friends. All speakers were native speakers of Czech, raised in Prague or in the region of Central Bohemia, and were between 19 and 26 years old. Every group of speakers consisted of one confederate, who was instructed to keep the conversations lively, and two speakers naive to the purposes of the recordings. The naive speakers were engaged in conversations for approximately 90 minutes, while the confederate joined them for approximately the last 72 minutes. The corpus was orthographically annotated by experienced transcribers and this orthographic transcription was aligned with the speech signal. In addition, the conversations were videotaped. This corpus can form the basis for all types of research on casual conversations in Czech, including phonetic research and research on how to improve automatic speech recognition. The corpus will be freely available

Abstract

In spontaneous conversations between adults, words are often pronounced with fewer segments or syllables than their citation forms. The question arises whether infant-directed speech also contains phonetic reduction. If so, infants would be presented with speech input that enables them to acquire reduced variants from an early age. This study compared speech directed at 11- and 12-month-old infants with adult-directed conversational speech and adult-directed read speech. In an acoustic study, 216 tokens of the Dutch words allemaal and helemaal from speech corpora were analyzed for duration, number of syllables, and vowel quality. In a perception study, adult participants rated these same materials for reduction and provided phonetic transcriptions. The results show that these two words are frequently reduced in infant-directed speech, and that their degree of reduction is comparable with conversational adult-directed speech. These findings suggest that lexical representations for reduced pronunciation variants can be acquired early in linguistic development

Abstract

This paper describes the pilot study of phonetic segmentation applied to Nijmegen Corpus of Casual Czech (NCCCz). This corpus contains informal speech of strong spontaneous nature which influences the character of produced speech at various levels. This work is the part of wider research related to the analysis of pronunciation reduction in such informal speech. We present the analysis of the accuracy of phonetic segmentation when canonical or reduced pronunciation is used. The achieved accuracy of realized phonetic segmentation provides information about general accuracy of proper acoustic modelling which is supposed to be applied in spontaneous speech recognition. As a byproduct of presented spontaneous speech segmentation, this paper also describes the created lexicon with canonical pronunciations of words in NCCCz, a tool supporting pronunciation check of lexicon items, and finally also a minidatabase of selected utterances from NCCCz manually labelled on phonetic level suitable for evaluation purposes

Abstract

This study examines how lexical frequency and planning problems can predict phonetic variability in the function word ‘the’ in conversational speech produced by non-native speakers of English. We examined 3180 tokens of ‘the’ drawn from English conversations between native speakers of Czech or Norwegian. Using regression models, we investigated the effect of following word frequency and disfluencies on three phonetic parameters: vowel duration, vowel quality, and consonant quality. Overall, the non-native speakers showed variation that is very similar to the variation displayed by native speakers of English. Like native speakers, Czech speakers showed an effect of frequency on vowel durations, which were shorter in more frequent word sequences. Both groups of speakers showed an effect of frequency on consonant quality: the substitution of another consonant for /ð/ occurred more often in the context of more frequent words. The speakers in this study also showed a native-like allophonic distinction in vowel quality, in which /ði/ occurs more often before vowels and /ðə/ before consonants. Vowel durations were longer in the presence of following disfluencies, again mirroring patterns in native speakers, and the consonant quality was more likely to be the target /ð/ before disfluencies, as opposed to a different consonant. The fact that non-native speakers show native-like sensitivity to lexical frequency and disfluencies suggests that these effects are consequences of a general, non-language-specific production mechanism governing language planning. On the other hand, the non-native speakers in this study did not show native-like patterns of vowel quality in the presence of disfluencies, suggesting that the pattern attested in native speakers of English may result from language-specific processes separate from the general production mechanisms

Abstract

This paper addresses the question how to compare reaction times computed by a computational model of speech comprehension with observed reaction times by participants. The question is based on the observation that reaction time sequences substantially differ per participant, which raises the issue of how exactly the model is to be assessed. Part of the variation in reaction time sequences is caused by the so-called local speed: the current reaction time correlates to some extent with a number of previous reaction times, due to slowly varying variations in attention, fatigue etc. This paper proposes a method, based on time series analysis, to filter the observed reaction times in order to separate the local speed effects. Results show that after such filtering the between-participant correlations increase as well as the average correlation between participant and model increases. The presented technique provides insights into relevant aspects that are to be taken into account when comparing reaction time sequences

Abstract

This study investigates whether listeners' familiarity with an intonation contour affects speech processing. In three experiments, Dutch participants heard Dutch sentences with normal intonation contours and with unfamiliar ones and performed word-monitoring, lexical decision, or semantic categorisation tasks (the latter two with cross-modal identity priming). The unfamiliar intonation contour slowed down participants on all tasks, which demonstrates that an unfamiliar intonation contour has a robust detrimental effect on speech processing. Since cross-modal identity priming with a lexical decision task taps into lexical access, this effect obtained in this task suggests that an unfamiliar intonation contour hinders lexical access. Furthermore, results from the semantic categorisation task show that the effect of an uncommon intonation contour is long-lasting and hinders subsequent processing. Hence, intonation not only contributes to utterance meaning (emotion, sentence type, and focus), but also affects crucial aspects of the speech comprehension process and is more important than previously thought.

Abstract

Paper presented at The 161th Meeting Acoustical Society of America, Seattle, Washington, 23-27 May 2011. Reduction in casual speech can create ambiguity, e.g., “he was” can sound like “he’s.” Before quotative “like” “so she’s/she was like…”, it was found that there is little accurate acoustic information about the distinction in the signal. This work examines what types of information acoustics of the target itself, speech rate, coarticulation, and syntax/semantics listeners use to recognize such reduced function words. We compare perception studies presenting the targets auditorily with varying amounts of context, presenting the context without the targets, and a visual study presenting context in written form. Given primarily discourse information visual or auditory context only, subjects are strongly biased toward past, reflecting the use of quotative “like” for reporting past speech. However, if the target itself is presented, the direction of bias reverses, indicating that listeners favor acoustic information within the target which is reduced, sounding like the shorter, present form over almost any other source of information. Furthermore, when the target is presented auditorily with surrounding context, the bias shifts slightly toward the direction shown in the orthographic or auditory-no-target experiments. Thus, listeners prioritize acoustic information within the target when present, even if that information is misleading, but they also take discourse information into account.

Abstract

This study presents an analysis of over 4000 tokens of words produced as variants with and without schwa in a French corpus of radio-broadcasted speech. In order to determine which of the many variables mentioned in the literature influence variant choice, 17 predictors were tested in the same analysis. Only five of these variables appeared to condition variant choice. The question of the processing stage, or locus, of this alternation process is also addressed in a comparison of the variables that predict variant choice with the variables that predict the acoustic duration of schwa in variants with schwa. Only two variables predicting variant choice also predict schwa duration. The limited overlap between the predictors for variant choice and for schwa duration, combined with the nature of these variables, suggest that the variants without schwa do not result from a phonetic process of reduction; that is, they are not the endpoint of gradient schwa shortening. Rather, these variants are generated early in the production process, either during phonological encoding or word-form retrieval. These results, based on naturally produced speech, provide a useful complement to on-line production experiments using artificial speech tasks.

Abstract

Words are often pronounced very differently in formal speech than in everyday conversations. In conversational speech, they may contain weaker segments, fewer sounds, and even fewer syllables. The English word yesterday, for instance, may be pronounced as [j epsilon integral eI]. This article forms an introduction to the phenomenon of reduced pronunciation variants and to the eight research articles in this issue on the characteristics, production, and comprehension of these variants. We provide a description of the phenomenon, addressing its high frequency of occurrence in casual conversations in various languages, the gradient nature of many reduction processes, and the intelligibility of reduced variants to native listeners. We also describe the relevance of research on reduced variants for linguistic and psychological theories as well as for applications in speech technology and foreign language acquisition. Since reduced variants occur more often in spontaneous than in formal speech, they are hard to study in the laboratory under well controlled conditions. We discuss the advantages and disadvantages of possible solutions, including the research methods employed in the articles in this special issue, based on corpora and experiments. This article ends with a short overview of the articles in this issue.

Abstract

This corpus study demonstrates that the realization of wordfinal /t/ in Dutch past-participles in various speech styles is affected by a word’s predictability and paradigmatic relative frequency. In particular, /t/s are shorter and more often absent if the two preceding words are more predictable. In addition, /t/s, especially in irregular verbs, are more reduced, the lower the verb’s lemma frequency relative to the past-participle’s frequency. Both effects are more pronounced in more spontaneous speech. These findings are expected if speech planning plays an important role in speech reduction. Index Terms: pronunciation variation, acoustic reduction, corpus research, word predictability, morphological decomposability

Abstract

This study investigates the prosodic conditioning of phonetic details which are candidate cues to phonological contrasts. German /b, d, g, p, t, k/ were examined in three prosodic positions. Lenis plosives /b, d, g/ were produced with less glottal vibration at larger prosodic boundaries, whereas their VOT showed no effect of prosody. VOT of fortis plosives /p, t, k/ decreased at larger boundaries, as did their burst intensity maximum. Vowels (when measured from consonantal release) following fortis plosives and lenis velars were shorter after larger boundaries. Closure duration, which did not contribute to the fortis/lenis contrast, was heavily affected by prosody. These results support neither of the hitherto proposed accounts of prosodic strengthening (Uniform Strengthening and Feature Enhancement). We propose a different account, stating that the phonological identity of speech sounds remains stable not only within, but also across prosodic positions (contrast-over-prosody hypothesis). Domain-initial strengthening hardly diminishes the contrast between prosodically weak fortis and strong lenis plosives.

Abstract

In spontaneous, conversational speech, words are often reduced compared to their citation forms, such that a word like yesterday may sound like [’jεsmall eshei]. The present chapter investigates such acoustic reduction. The study of reduction needs large corpora that are transcribed phonetically. The first part of this chapter describes an automatic transcription procedure used to obtain such a large phonetically transcribed corpus of Dutch spontaneous dialogues, which is subsequently used for the investigation of acoustic reduction. First, the orthographic transcriptions were adapted for automatic processing. Next, the phonetic transcription of the corpus was created by means of a forced alignment using a lexicon with multiple pronunciation variants per word. These variants were generated by applying phonological and reduction rules to the canonical phonetic transcriptions of the words. The second part of this chapter reports the results of a quantitative analysis of reduction in the corpus on the basis of the generated transcriptions and gives an inventory of segmental reductions in standard Dutch. Overall, we found that reduction is more pervasive in spontaneous Dutch than previously documented.

Abstract

This paper presents a method to quantify the spectral characteristics of reduction in speech. Hämäläinen et al. (2009) proposes a measure of spectral reduction which is able to predict a substantial amount of the variation in duration that linguistically motivated variables do not account for. In this paper, we continue studying acoustic reduction in speech by developing a new acoustic measure of reduction, based on local manifold structure in speech. We show that this measure yields significantly improved statistical models for predicting variation in duration.

Abstract

The present study compares the realization of intervocalic voiceless stops and vowels surrounded by voiceless stops in conversational Spanish and French. Our data reveal significant differences in how these segments are realized in each language. Spanish voiceless stops tend to have shorter stop closures, display incomplete closures more often, and exhibit more voicing than French voiceless stops. As for vowels, more cases of complete devoicing and greater degrees of partial devoicing were found in French than in Spanish. Moreover, all French vowel types exhibit significantly lower F1 values than their Spanish counterparts. These findings indicate that the extent of reduction that a segment type can undergo in conversational speech can vary significantly across languages. Language differences in coarticulatory strategies and “base-of-articulation” are discussed as possible causes of our observations.

Abstract

This study investigates the reduction of vowel /e/ in the French word c’était /setε/ ‘it was’. This reduction phenomenon appeared to be highly frequent, as more than half of the occurrences of this word in a corpus of casual French contained few or no acoustic traces of a vowel between [s] and [t]. All our durational analyses clearly supported a categorical absence of vowel /e/ in a subset of c’était tokens. This interpretation was also supported by our finding that the occurrence of complete elision and [e] duration in non-elision tokens were conditioned by different factors. However, spectral measures were consistent with the possibility that a highly reduced /e/ vowel is still present in elision tokens in spite of the durational evidence for categorical elision. We discuss how these findings can be reconciled, and conclude that acoustic analysis of uncontrolled materials can provide valuable information about the mechanisms underlying reduction phenomena in casual speech.

Abstract

This study investigates the lifespans of lexical traces for novel morphologically complex words. In two visual lexical decision experiments, a neologism was either primed by itself or by its stem. The target occurred 40 trials after the prime (Experiments 1 & 2), after a 12 hour delay (Experiment 1), or after a one week delay (Experiment 2). Participants recognized neologisms more quickly if they had seen them before in the experiment. These results show that memory traces for novel morphologically complex words already come into existence after a very first exposure and that they last for at least a week. We did not find evidence for a role of sleep in the formation of memory traces. Interestingly, Base Frequency appeared to play a role in the processing of the neologisms also when they were presented a second time and had their own memory traces.

Abstract

Listeners require context to understand the highly reduced words that occur in casual speech. The present study reports four auditory lexical decision experiments in which the role of semantic context in the comprehension of reduced versus unreduced speech was investigated. Experiments 1 and 2 showed semantic priming for combinations of unreduced, but not reduced, primes and low-frequency targets. In Experiment 3, we crossed the reduction of the prime with the reduction of the target. Results showed no semantic priming from reduced primes, regardless of the reduction of the targets. Finally, Experiment 4 showed that reduced and unreduced primes facilitate upcoming low-frequency related words equally if the interstimulus interval is extended. These results suggest that semantically related words need more time to be recognized after reduced primes, but once reduced primes have been fully (semantically) processed, these primes can facilitate the recognition of upcoming words as well as do unreduced primes.