Publications

Abstract

This doctoral dissertation investigated the learning of auditory categories by applying insights from child language learning, visual category learning and phonetic categorization by adults. The experiments in chapter 2 concern supervised learning of multidimensional non-speech categories varying in two dimensions: duration and a non speech analogue of formant frequency. In experiment 1, one dimension of variation determined category membership, variation in the other dimension was irrelevant. Listeners quickly learned to categorize according to this distinction. In experiment 2, both dimensions needed to be combined to categorize correctly. Performance was much lower, but most listeners succeeded in this task. However, in a maintenance phase without feedback or distributional information, listeners reverted to a unidimensional solution. In a maintenance phase with distributional information, listeners did use both dimensions correctly, arguing for the importance of distributional information in (auditory) category acquisition. In chapter 3, the listeners had to classify the same categories, but without feedback. In experiment 1, listeners succeeded to discover the relevant dimension of variation (and ignore the irrelevant one) without feedback. Much of this learning was lost in the maintenance phase, especially for the dimension formant frequency. With two relevant dimensions (Experiment 2), listeners were not able to learn to use both dimensions and reverted to a unidimensional solution. Chapter 4 applied the paradigms of chapter 2 and 3 to the learning of speech categories. Spanish native listeners learned Dutch vowel contrast with one relevant dimension of variation. With feedback, learning was swift, although was not well maintained without feedback or distributional information. Without feedback, Spanish listeners reverted to the dimensions best known in their native phonology, formant frequency, even when distributional information pointed to duration. The results are discussed in chapter 5. The implications for models of speech acquisition are discussed.

Abstract

Distributional learning is almost certainly involved in the human acquisition of phonetic categories. Because speech is inherently a multidimensional signal, learning phonetic categories entails multidimensional learning. Yet previous studies of auditory category learning have shown poor maintenance of learned multidimensional categories. Two experiments explored ways to improve maintenance: by increasing the costs associated with applying a unidimensional strategy; by providing additional information about the category structures; and by giving explicit instructions on how to categorize. Only with explicit instructions were categorization strategies maintained in a maintenance phase without supervision or distributional information.

Abstract

The standard for the analysis of distorted voices is perceptual rating of read-out texts or spontaneous speech. Automatic voice evaluation, however, is usually done on stable sections of sustained vowels. In this paper, text-based and established vowel-based analysis are compared with respect to their ability to measure hoarseness and its subclasses. 73 hoarse patients (48.3±16.8 years) uttered the vowel /e/ and read the German version of the text “The North Wind and the Sun”. Five speech therapists and physicians rated roughness, breathiness, and hoarseness according to the German RBH evaluation scheme. The best human-machine correlations were obtained for measures based on the Cepstral Peak Prominence (CPP; up to |r | = 0.73). Support Vector Regression (SVR) on CPP-based measures and prosodic features improved the results further to r ≈0.8 and confirmed that automatic voice evaluation should be performed on a text recording.

Abstract

Child songs are a great source for linguistic learning. Here we explore whether children can acquire novel words in a second language by playing a game featuring child songs in a playhouse. We present data from three studies that serve as scientific proof for the functionality of one game of the playhouse: the Energy Center. For this game, three hand-bikes were mounted on a panel. When children start moving the hand-bikes, child songs start playing simultaneously. Once the children produce enough energy with the hand-bikes, the songs are additionally accompanied with the sounds of musical instruments. In our studies, children executed a picture-selection task to evaluate whether they acquired new vocabulary from the songs presented during the game. Two of our studies were run in the field, one at a Dutch and one at an Indian pre-school. The third study features data from a more controlled laboratory setting. Our results partly confirm that the Energy Center is a successful means to support vocabulary acquisition in a second language. More research with larger sample sizes and longer access to the Energy Center is needed to evaluate the overall functionality of the game. Based on informal observations at our test sites, however, we are certain that children do pick up linguistic content from the songs during play, as many of the children repeat words and phrases from songs they heard. We will pick up upon these promising observations during future studies

Abstract

This study investigated whether a shadowing task can provide insights in the nature of reduction processes that are typical of casual speech. We focused on the shortening and presence versus absence of schwa and /t/ in Dutch past participles. Results showed that the absence of these segments was affected by the same variables as their shortening, suggesting that absence mostly resulted from extreme gradient shortening. This contrasts with results based on recordings of spontaneous conversations. We hypothesize that this difference is due to non-casual fast speech elicited by a shadowing task.

Abstract

The Possible-Word Constraint (PWC; Norris, McQueen, Cutler, and Butterfield 1997) suggests that segmentation of continuous speech operates with a universal constraint that feasible words should contain a vowel. Single consonants, because they do not constitute syllables, are treated as non-viable residues. Two word-spotting experiments are reported that investigate whether the PWC really is a language-universal principle. According to the PWC, Slovak listeners should, just like Germans, be slower at spotting words in single consonant contexts (not feasible words) as compared to syllable contexts (feasible words)—even if single consonants can be words in Slovak. The results confirm the PWC in German but not in Slovak.

Abstract

Syntactic parsers and generators need highquality grammars of coordination and coordinate ellipsis—structures that occur very frequently but are much less well understood theoretically than many other domains of grammar. Modern grammars of coordinate ellipsis are based nearly exclusively on linguistic judgments (intuitions). The extent to which grammar rules based on this type of empirical evidence generate all and only the structures in text corpora, is unknown. As part of a project on the development of a grammar and a generator for coordinate ellipsis in German, we undertook an extensive exploration of the TIGER treebank—a syntactically annotated corpus of about 50,000 newspaper sentences. We report (1) frequency data for the various patterns of coordinate ellipsis, and (2) several rarely (but regularly) occurring ‘fringe deviations’ from the intuition-based rules for several ellipsis types. This information can help improve parser and generator performance.

Abstract

Certain categories of language learners need feedback on the grammatical structure of sentences they wish to produce. In contrast with the usual NLP approach to this problem—parsing student-generated texts—we propose a generation-based approach aiming at preventing errors (“scaffolding”). In our ICALL system, students construct sentences by composing syntactic trees out of lexically anchored “treelets” via a graphical drag&drop user interface. A natural-language generator computes all possible grammatically well-formed sentences entailed by the student-composed tree, and intervenes immediately when the latter tree does not belong to the set of well-formed alternatives. Feedback is based on comparisons between the student-composed tree and the well-formed set. Frequently occurring errors are handled in terms of “malrules.” The system (implemented in JAVA and C++) currently focuses constituent order in German as L2.

Abstract

The potential of sentence generators as engines in Intelligent Computer-Assisted Language Learning and teaching (ICALL) software has hardly been explored. We sketch the prototype of COMPASS, a system that supports integrated writing and grammar curricula for 10 to 14 year old elementary or secondary schoolers. The system enables first- or second-language teachers to design controlled writing exercises, in particular of the “sentence combining” variety. The system includes facilities for error diagnosis and on-line feedback. Syntactic structures built by students or system can be displayed as easily understood phrase-structure or dependency trees, adapted to the student’s level of grammatical knowledge. The heart of the system is a specially designed generator capable of lexically guided sentence generation, of generating syntactic paraphrases, and displaying syntactic structures visually.

Abstract

This thesis elaborates on a methodological approach to reliably infer cognitive preferences in an extinct evolutionary ancestor of modern humans. In attempts to understand cognitive evolution, humans have been compared to capuchin monkeys, tamarins, and chimpanzees to name but a few. But comparisons between humans and one other, maybe even distantly related primate, as interesting as they might be, will not tell us anything about an evolutionary ancestor to humans. To put it bluntly: None of the living primates, not even chimpanzees, are a human ancestor. With that in mind, we can still use a comparative approach to gain information about our evolutionary ancestors, as long as we are careful about whom we compare with whom. If a certain trait exists in all genera of a phylogenetic clade, it was most likely present in their common ancestor. The great apes are such a clade (Pongo, Gorilla, Pan and Homo). It follows that, if members of all great ape genera shared a particular cognitive preference or ability, it is most likely part of the evolutionary inheritance of the clade at least ever since their last common ancestor, and therefore also an evolutionarily old, inherited cognitive default in humans. This thesis contains studies comparing all 4 extant Hominid genera, including humans of 4 different age-groups and 2 different cultures. Results show that all great apes do indeed share some cognitive preferences, which they most likely inherited from an evolutionary ancestor. Additionally, human cognitive preferences can change away from such an inherited predisposition given ontogenetic factors, and are at least in part variably adaptable to cultural circumstance.

Abstract

Prediction in language has traditionally been studied using
simple designs in which neural responses to expected
and unexpected words are compared in a categorical
fashion. However, these designs have been contested
as being ‘prediction encouraging’, potentially exaggerating
the importance of prediction in language understanding.
A few recent studies have begun to address
these worries by using model-based approaches to probe
the effects of linguistic predictability in naturalistic stimuli
(e.g. continuous narrative). However, these studies
so far only looked at very local forms of prediction, using
models that take no more than the prior two words into
account when computing a word’s predictability. Here,
we extend this approach using a state-of-the-art neural
language model that can take roughly 500 times longer
linguistic contexts into account. Predictability estimates
fromthe neural network offer amuch better fit to EEG data
from subjects listening to naturalistic narrative than simpler
models, and reveal strong surprise responses akin to
the P200 and N400. These results show that predictability
effects in language are not a side-effect of simple designs,
and demonstrate the practical use of recent advances
in AI for the cognitive neuroscience of language.

Abstract

This paper reports on findings from an elicited production task with German 5-year-old children, investigating their use of intonation to mark information status of discourse referents. In line with findings for adults, new referents were preferably marked by H* and L+H*; textually given referents were mainly deaccented. Accessible referents (whose first mentions were less recent) were mostly accented, and predominantly also realised with H* and L+H*, showing children’s sensitivity to recency of mention. No evidence for the consistent use of a special ‘accessibility accent’ H+L* (as has been proposed for adult German) was found.

Abstract

Co-speech gestures are an integral part of human face-to-face communication, but little is known about how pragmatic factors influence our comprehension of those gestures. The present study investigates how different types of recipients process iconic gestures in a triadic communicative situation. Participants (N = 32) took on the role of one of two recipients in a triad and were presented with 160 video clips of an actor speaking, or speaking and gesturing. Crucially, the actor’s eye gaze was manipulated in that she alternated her gaze between the two recipients. Participants thus perceived some messages in the role of addressed recipient and some in the role of unaddressed recipient. In these roles, participants were asked to make judgements concerning the speaker’s messages. Their reaction times showed that unaddressed recipients did comprehend speaker’s gestures differently to addressees. The findings are discussed with respect to automatic and controlled processes involved in gesture comprehension.

Abstract

Thesaurus alignment plays an important role in realising efficient access to heterogeneous Cultural Heritage data. Current ontology alignment techniques, however, provide only limited value for such access as they consider little if any requirements from realistic use cases or application scenarios. In this paper, we focus on two real-world scenarios in a library context: thesaurus merging and book re-indexing. We identify their particular requirements and describe our approach of deploying and evaluating thesaurus alignment techniques in this context. We have applied our approach for the Ontology Alignment Evaluation Initiative, and report on the performance evaluation of participants’ tools wrt. the application scenario at hand. It shows that evaluations of tools requires significant effort, but when done carefully, brings many benefits.

Abstract

Thesaurus alignment is important for efficient access to heterogeneous Cultural Heritage data. Current ontology alignment techniques provide solutions, but with limited value in practice, because the requirements from usage scenarios are rarely taken in account. In this paper, we start from particular requirements for book re-indexing and investigate possible ways of developing, deploying and evaluating thesaurus alignment techniques in this context. We then compare different aspects of this scenario with others from a more general perspective.

Abstract

In this study we investigated to what extent a meaningful sentence context facilitates spoken word processing in young and older listeners if listening is made taxing by time-compressing the speech. Even though elderly listeners have been shown to benefit more from sentence context in difficult listening conditions than young listeners, time compression of speech may interfere with semantic comprehension, particularly in older listeners because of cognitive slowing. The results of a target detection experiment showed that, unlike young listeners who showed facilitation by context at both rates, elderly listeners showed context facilitation at the intermediate, but not at the fastest rate. This suggests that semantic interpretation lags behind target identification.

Abstract

Listeners use lexical knowledge to adjust their prelexical representations of speech sounds in response to the idiosyncratic pronunciations of particular speakers. We used an exposure-test paradigm to investigate whether this type of perceptual learning transfers across syllabic positions. No significant learning effect was found in Experiment 1, where exposure sounds were onsets and test sounds were codas. Experiments 2-4 showed that there was no learning even when both exposure and test sounds were onsets. But a trend was found when exposure sounds were codas and test sounds were onsets (Experiment 5). This trend was smaller than the robust effect previously found for the coda-to-coda case. These findings suggest that knowledge about idiosyncratic pronunciations may be position specific: Knowledge about how a speaker produces sounds in one position, if it can be acquired at all, influences perception of sounds in that position more strongly than of sounds in another position.

Abstract

Words repeated in the same voice are better recognized than when they are repeated in a different voice. Such findings have been taken as evidence for the storage of talker-specific lexical episodes. But results on perceptual learning suggest that talker-specific adjustments concern sublexical representations. This study thus investigates whether voice-specific repetition effects in auditory lexical decision are lexical or sublexical. The same critical set of items in Block 2 were, depending on materials in Block 1, either same-voice or different-voice word repetitions, new words comprising re-orderings of phonemes used in the same voice in Block 1, or new words with previously unused phonemes. Results show a benefit for words repeated by the same talker, and a smaller benefit for words consisting of phonemes repeated by the same talker. Talker-specific information thus appears to influence word recognition at multiple representational levels.

Abstract

Listeners use suprasegmental auditory lexical stress information to resolve the competition words engage in during spoken-word recognition. The present study investigated whether (a) visual speech provides lexical stress information, and, more importantly, (b) whether this visual lexical stress information is used to resolve lexical competition. Dutch word pairs that differ in the lexical stress realization of their first two syllables, but not segmentally (e.g., 'OCtopus' and 'okTOber'; capitals marking primary stress) served as auditory-only, visual-only, and audiovisual speech primes. These primes either matched (e.g., 'OCto-'), mismatched (e.g., 'okTO-'), or were unrelated to (e.g., 'maCHI-') a subsequent printed target (octopus), which participants had to make a lexical decision to. To the degree that visual speech contains lexical stress information, lexical decisions to printed targets should be modulated through the addition of visual speech. Results show, however, no evidence for a role of visual lexical stress information in audiovisual spoken-word recognition.

Abstract

Adult-to-child interactions are often characterized by prosodically-exaggerated speech accompanied by visually captivating co-speech gestures. In a series of adult studies, we have shown that these gestures are linked in a sophisticated manner to the prosodic structure of adults' utterances. In the current study, we use the Preferential Looking Paradigm to demonstrate that two-year-olds can use the alignment of these gestures to speech to deduce the meaning of words.

Abstract

This study investigates effects of prosodic factors (prominence, boundary) on coarticulatory Vnasalization in Australian English (AusE) in CVN and NVC in comparison to those in American English
(AmE). As in AmE, prominence was found to
lengthen N, but to reduce V-nasalization, enhancing N’s nasality and V’s orality, respectively (paradigmatic contrast enhancement). But the prominence effect in CVN was more robust than that in AmE. Again similar to findings in AmE, boundary
induced a reduction of N-duration and V-nasalization phrase-initially (syntagmatic contrast enhancement), and increased the nasality of both C and V phrasefinally.
But AusE showed some differences in terms
of the magnitude of V nasalization and N duration. The results suggest that the linguistic contrast enhancements underlie prosodic-structure modulation of coarticulatory V-nasalization in
comparable ways across dialects, while the fine phonetic detail indicates that the phonetics-prosody interplay is internalized in the individual dialect’s phonetic grammar.

Abstract

The ”download first, then process paradigm” is still the predominant working method amongst the research community. The web-based paradigm, however, offers many advantages from a tool development and data management perspective as they allow a quick adaptation to changing research environments. Moreover, new ways of combining tools and data are increasingly becoming available and will eventually enable a true web-based workflow approach, thus challenging the ”download first, then process” paradigm. The necessary infrastructure for managing, exploring and enriching language resources via the Web will need to be delivered by projects like CLARIN and DARIAH

Abstract

In this paper, we describe a unifying approach to tackle data heterogeneity issues for lexica and related resources. We present LEXUS, our software that implements the Lexical Markup Framework (LMF) to uniformly describe and manage lexica of different structures. LEXUS also makes use of a central Data Category Registry (DCR) to address terminological issues with regard to linguistic concepts as well as the handling of working and object languages. Finally, we report on ViCoS, a LEXUS extension, providing support for the definition of arbitrary semantic relations between lexical entries or parts thereof.

Abstract

To achieve true interoperability for valuable linguistic resources different levels of variation need to be addressed. ISO Technical Committee 37, Terminology and other language and content resources, is developing a Data Category Registry. This registry will provide a reusable set of data categories. A new implementation, dubbed ISOcat, of the registry is currently under construction. This paper shortly describes the new data model for data categories that will be introduced in this implementation. It goes on with a sketch of the standardization process. Completed data categories can be reused by the community. This is done by either making a selection of data categories using the ISOcat web interface, or by other tools which interact with the ISOcat system using one of its various Application Programming Interfaces. Linguistic resources that use data categories from the registry should include persistent references, e.g. in the metadata or schemata of the resource, which point back to their origin. These data category references can then be used to determine if two or more resources share common semantics, thus providing a level of interoperability close to the source data and a promising layer for semantic alignment on higher levels

Abstract

Word segmentation, or detecting word boundaries in continuous speech, is not an easy task. Spoken language does not contain silences to indicate word boundaries and words partly overlap due to coarticalution. Still, adults listening to their native language perceive speech as individual words. They are able to combine different distributional cues in the language, such as the statistical distribution of sounds and metrical cues, with lexical information, to efficiently detect word boundaries. Infants in the first year of life do not command these cues. However, already between seven and ten months of age, before they know word meaning, infants learn to segment words from speech. This important step in language acquisition is the topic of this dissertation. In chapter 2, the first Event Related Brain Potential (ERP) study on word segmentation in Dutch ten-month-olds is discussed. The results show that ten-month-olds can already segment words with a strong-weak stress pattern from speech and they need roughly the first half of a word to do so. Chapter 3 deals with segmentation of words beginning with a weak syllable, as a considerable number of words in Dutch do not follow the predominant strong-weak stress pattern. The results show that ten-month-olds still largely rely on the strong syllable in the language, and do not show an ERP response to the initial weak syllable. In chapter 4, seven-month-old infants' segmentation of strong-weak words was studied. An ERP response was found to strong-weak words presented in sentences. However, a behavioral response was not found in an additional Headturn Preference Procedure study. There results suggest that the ERP response is a precursor to the behavioral response that infants show at a later age.

Abstract

The present study investigates the influence of prosodic structure on the fine-grained phonetic details of German plosives which also cue the phonological fortis-lenis contrast. Closure durations were found to be longer at higher prosodic boundaries. There was also less glottal vibration in lenis plosives at higher prosodic boundaries. Voice onset time in lenis plosives was not affected by prosody. In contrast, for the fortis plosives VOT decreased at higher boundaries, as did the maximal intensity of the release. These results demonstrate that the effects of prosody on different phonetic cues can go into opposite directions, but are overall constrained by the need to maintain phonological contrasts. While prosodic effects on some cues are compatible with a ‘fortition’ account of prosodic strengthening or with a general feature enhancement explanation, the effects on others enhance paradigmatic contrasts only within a given prosodic position.

Abstract

In this study, we explored methods for linking Chinese and English sense inventories using two opposing approaches: creating links (1) bottom-up: by starting at the finer-grained sense level then proceeding to the verb subcategorization frames and (2) top-down: by starting directly with the more coarse-grained frame levels. The sense inventories for linking include pre-existing corpora, such as English Propbank (Palmer, Gildea, and Kingsbury, 2005), Chinese Propbank (Xue and Palmer, 2004) and English WordNet (Fellbaum, 1998) and newly created corpora, the English and Chinese Sense Inventories from DARPA-GALE OntoNotes. In the linking task, we selected a group of highly frequent and polysemous communication verbs, including say, ask, talk, and speak in English, and shuo, biao-shi, jiang, and wen in Chinese. We found that with the bottom-up method, although speakers of both languages agreed on the links between senses, the subcategorization frames of the corresponding senses did not match consistently. With the top-down method, if the verb frames match in both languages, their senses line up more quickly to each other. The results indicate that the top-down method is more promising in linking English and Chinese sense inventories.

Abstract

In different fields of the humanities annotations of multimodal resources are a necessary component of the research workflow. Examples include linguistics, psychology, anthropology, etc. However, creation of those annotations is a very laborious task, which can take 50 to 100 times the length of the annotated media, or more. This can be significantly improved by applying innovative audio and video processing algorithms, which analyze the recordings and provide automated annotations. This is the aim of the AVATecH project, which is a collaboration of the Max Planck Institute for Psycholinguistics (MPI) and the Fraunhofer institutes HHI and IAIS. In this paper we present a set of results of automated annotation together with an evaluation of their quality.

Abstract

Digital processing of medical images has helped physicians and patients during past years by allowing examination and diagnosis on a very precise level. Nowadays possibly the biggest deal of support it can offer for modern healthcare is the use of high performance computing architectures to treat the huge amounts of data that can be collected by modern acquisition devices. This paper presents a parallel processing implementation of an image segmentation algorithm that operates on a computer cluster equipped with 10 processing units. Thanks to well-organized distribution of the workload we manage to significantly shorten the execution time of the developed algorithm and reach a performance gain very close to linear.

Abstract

Introduction Human spoken communication is multimodal, i.e. it encompasses both speech and gesture. Acoustic properties of voice, body movements, facial expression, etc. are an inherent and meaningful part of spoken interaction; they can provide attitudinal, grammatical and semantic information. In the recent years interest in audio-visual corpora has been rising rapidly as they enable investigation of different communicative modalities and provide more holistic view on communication (Kipp et al. 2009). Moreover, for some languages such corpora are the only available resource, as is the case for endangered languages for which no written resources exist.

Abstract

In this paper we describe audio and video processing algorithms that are developed in the scope of AVATecH project. The purpose of these algorithms is to shorten the time taken by manual annotation of audio and video recordings by extracting features from media files and creating semi-automated annotations. We show that the use of such supporting algorithms can shorten the annotation time to 30-50% of the time necessary to perform a fully manual annotation of the same kind.

Abstract

Verb learning is important for young children. While most
previous research has focused on linguistic and conceptual
challenges in early verb learning (e.g. Gentner, 1982, 2006),
the present paper examined early verb learning at the
attentional level and quantified the input for early verb learning
by measuring verb-action co-occurrence statistics in parent-
child interaction from the learner’s perspective. To do so, we
used head-mounted eye tracking to record fine-grained
multimodal behaviors during parent-infant joint play, and
analyzed parent speech, parent and infant action, and infant
attention at the moments when parents produced verb labels.
Our results show great variability across different action verbs,
in terms of frequency of verb utterances, frequency of
corresponding actions related to verb meanings, and infants’
attention to verbs and actions, which provide new insights on
why some verbs are harder to learn than others.

Abstract

Young children demonstrate the ability to make inferences about the preferences of other agents based on their choices. However, there exists no overarching account of what children are doing when they learn about preferences or how they use that knowledge. We use a rational model of preference learning, drawing on ideas from economics and computer science, to explain the behavior of children in several recent experiments. Specifically, we show how a simple econometric model can be extended to capture two- to four-year-olds’ use of statistical information in inferring preferences, and their generalization of these preferences.

Abstract

We examined how communicators can switch between speaker and listener role with such accurate timing. During conversations, the majority of role transitions happens with a gap or overlap of only a few hundred milliseconds. This suggests that listeners can predict when the turn of the current speaker is going to end. Our hypothesis is that listeners know when a turn ends because they know how it ends. Anticipating the last words of a turn can help the next speaker in predicting when the turn will end, and also in anticipating the content of the turn, so that an appropriate response can be prepared in advance. We used the stimuli material of an earlier experiment (De Ruiter, Mitterer & Enfield, 2006), in which subjects were listening to turns from natural conversations and had to press a button exactly when the turn they were listening to ended. In the present experiment, we investigated if the subjects can complete those turns when only an initial fragment of the turn is presented to them. We found that the subjects made better predictions about the last words of those turns that had more accurate responses in the earlier button press experiment.

Abstract

Continuous Bag of Words (CBOW) is a powerful text embedding method. Due to its strong capabilities to encode word content, CBOW embeddings perform well on a wide range of downstream tasks while being efficient to compute. However, CBOW is not capable of capturing the word order. The reason is that the computation of CBOW's word embeddings is commutative, i.e., embeddings of XYZ and ZYX are the same. In order to address this shortcoming, we propose a
learning algorithm for the Continuous Matrix Space Model, which we call Continual Multiplication of Words (CMOW). Our algorithm is an adaptation of word2vec, so that it can be trained on large quantities of unlabeled text. We empirically show that CMOW better captures linguistic properties, but it is inferior to CBOW in memorizing word content. Motivated by these findings, we propose a hybrid model that combines the strengths of CBOW and CMOW. Our results show that the hybrid CBOW-CMOW-model retains CBOW's strong ability to memorize word content while at the same time substantially improving its ability to encode other linguistic information by 8%. As a result, the hybrid also performs better on 8 out of 11 supervised downstream tasks with an average improvement of 1.2%.

Abstract

Scholars disagree about the extent to which language can tell us
about conceptualisation of the world. Some believe that language
is a direct window onto concepts: Having a word ‘‘bird’’, ‘‘table’’ or
‘‘sour’’ presupposes the corresponding underlying concept, BIRD,
TABLE, SOUR. Others disagree. Words are thought to be uninformative,
or worse, misleading about our underlying conceptual representations;
after all, our mental worlds are full of ideas that we
struggle to express in language. How could this be so, argue sceptics,
if language were a direct window on our inner life? In this presentation,
I consider what language can tell us about the
conceptualisation of taste. By considering linguistic data from
twenty unrelated cultures – varying in subsistence mode (huntergatherer
to industrial), ecological zone (rainforest jungle to desert),
dwelling type (rural and urban), and so forth – I argue any single language is, indeed, impoverished about what it can reveal about
taste. But recurrent lexicalisation patterns across languages can
provide valuable insights about human taste experience. Moreover,
language patterning is part of the data that a good theory of taste
perception has to be answerable for. Taste researchers, therefore,
cannot ignore the crosslinguistic facts.

Abstract

Studies have claimed that blind people’s spatial representations are different from sighted people, and blind people display superior auditory processing. Due to the nature of auditory and haptic information, it has been proposed that blind people have spatial representations that are more sequential than sighted people. Even the temporary loss of sight—such as through blindfolding—can affect spatial representations, but not much research has been done on this topic. We compared blindfolded and sighted people’s linguistic spatial expressions and non-linguistic localization accuracy to test how blindfolding affects the representation of path in auditory motion events. We found that blindfolded people were as good as sighted people when localizing simple sounds, but they outperformed sighted people when localizing auditory motion events. Blindfolded people’s path related speech also included more sequential, and less holistic elements. Our results indicate that even temporary loss of sight influences spatial representations of auditory motion events

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

This paper presents lexical statistics on the pattern of occurrence of words embedded in other words. We report the results of an analysis of 25000 words, varying in length from two to six syllables, extracted from a phonetically-coded English dictionary (The Longman Dictionary of Contemporary English). Each syllable, and each string of syllables within each word was checked against the dictionary. Two analyses are presented: the first used a complete list of polysyllables, with look-up on the entire dictionary; the second used a sublist of content words, counting only embedded words which were themselves content words. The results have important implications for models of human speech recognition. The efficiency of these models depends, in different ways, on the number and location of words within words.

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

Listeners use lexical knowledge to judge what speech sounds they heard. I investigated whether such lexical influences are truly top-down or just reflect a merging of perceptual and lexical constraints. This is achieved by testing whether the lexically determined identity of a phone exerts the appropriate context effects on surrounding phones. The current investigations focuses on compensation for coarticulation in vowel-fricative sequences, where the presence of a rounded vowel (/y/ rather than /i/) leads fricatives to be perceived as /s/ rather than //. This results was consistently found in all three experiments. A vowel was also more likely to be perceived as rounded /y/ if that lead listeners to be perceive words rather than nonwords (Dutch: meny, English id. vs. meni nonword). This lexical influence on the perception of the vowel had, however, no consistent influence on the perception of following fricative.

Abstract

We investigated perception of words with reduced word-final /t/ using an adapted eyetracking paradigm. Dutch listeners followed spoken instructions to click on printed words which were accompanied on a computer screen by simple shapes (e.g., a circle). Targets were either above or next to their shapes, and the shapes uniquely identified the targets when the spoken forms were ambiguous between words with or without final /t/ (e.g., bult, bump, vs. bul, diploma). Analysis of listeners’ eye-movements revealed, in contrast to earlier results, that listeners use the following segmental context when compensating for /t/-reduction. Reflecting that /t/-reduction is more likely to occur before bilabials, listeners were more likely to look at the /t/-final words if the next word’s first segment was bilabial. This result supports models of speech perception in which prelexical phonological processes use segmental context to modulate word recognition.

Abstract

To assess the reality of phonological features in language processing (vs. language description), one needs to specify the distinctive claims of distinctive-feature theory. Two of the more farreaching claims are compositionality and generalizability. I will argue that there is some evidence for the first and evidence against the second claim from a recent behavioral paradigm. Highlighting the contribution of a behavioral paradigm also counterpoints the use of brain measures as the only way to elucidate what is "real for the brain". The contributions of the speakers exemplify how brain measures can help us to understand the reality of phonological features in language processing. The evidence is, however, not convincing for a) the claim for underspecification of phonological features—which has to deal with counterevidence from behavioral as well as brain measures—, and b) the claim of position independence of phonological features.

Abstract

Words are reduced in spontaneous speech. If reductions are constrained by functional (i.e., perception and production) constraints, they should not be arbitrary. This hypothesis was tested by examing the pronunciations of high- to mid-frequency words in a Dutch and a German spontaneous speech corpus. In logistic-regression models the "reduction likelihood" of a phoneme was predicted by fixed-effect predictors such as position within the word, word length, word frequency, and stress, as well as random effects such as phoneme identity and word. The models for Dutch and German show many communalities. This is in line with the assumption that similar functional constraints influence reductions in both languages.

Abstract

The anterior lordosis of the cervical spine is thought
to contribute to pitch (fo) production by influencing
cricoid rotation as a function of larynx height. This
study examines the matter of inter-individual
variation in cervical spine shape and whether this has
an influence on how fo is produced along increasing
or decreasing scales, using the ArtiVarK dataset,
which contains real-time MRI pitch production data.
We find that the cervical spine actively participates in
fo production, but the amount of displacement
depends on individual shape. In general, anterior
spine motion (tending toward cervical lordosis)
occurs for low fo, while posterior movement (tending
towards cervical kyphosis) occurs for high fo.

Abstract

Studies have shown that listeners segmenting unfamiliar languages transfer native-language (L1) segmentation cues. These studies, however, conflated L1 and recent linguistic exposure. The present study investigates the relative influences of L1 and recent linguistic exposure on the use of prosodic cues for segmenting an artificial language (AL). Participants were L1-French listeners, high-proficiency L2-French L1-English listeners, and L1-English listeners without functional knowledge of French. The prosodic cue assessed was F0 rise, which is word-final in French, but in English tends to be word-initial. 30 participants heard a 20-minute AL speech stream with word-final boundaries marked by F0 rise, and decided in a subsequent listening task which of two words (without word-final F0 rise) had been heard in the speech stream. The analyses revealed a marginally significant effect of L1 (all listeners) and, importantly, a significant effect of recent linguistic exposure (L1-French and L2-French listeners): accuracy increased with decreasing time in the US since the listeners’ last significant (3+ months) stay in a French-speaking environment. Interestingly, no effect of L2 proficiency was found (L2-French listeners).

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

Language resources can be divided into structural resources treating phonology, morphosyntax, semantics etc. and resources treating the social, demographic, ethnic, political context. A third type are meta-resources, like bibliographies, which provide access to the resources of the first two kinds. This poster will present the Glottolog/Langdoc project, a comprehensive bibliography providing web access to 180k bibliographical records to (mainly) low visibility resources from low-density languages. The resources are annotated for macro-area, content language, and document type and are available in XHTML and RDF.

Abstract

We report two experiments in which vowels and steady-state consonants served as targets in a speeded detection task. In the first experiment, two vowels were compared with one voiced and once unvoiced fricative. Response times (RTs) to the vowels were longer than to the fricatives. The error rate was higher for the consonants. Consonants in word-final position produced the shortest RTs, For the vowels, RT correlated negatively with target duration. In the second experiment, the same two vowel targets were compared with two nasals. This time there was no significant difference in RTs, but the error rate was still significantly higher for the consonants. Error rate and length correlated negatively for the vowels only. We conclude that RT differences between phonemes are independent of vocalic or consonantal status. Instead, we argue that the process of phoneme detection reflects more finely grained differences in acoustic/articulatory structure within the phonemic repertoire.

Abstract

Frontotemporal lobar degeneration (FTLD) is a group of neurodegenerative conditions characterised by focal frontal and/or temporal lobe atrophy. Patients develop a range of cognitive and behavioural abnormalities, including prominent difficulties in comprehending and expressing emotions, with significant clinical and social consequences. Here we report a systematic prospective analysis of emotion processing in different input modalities in patients with FTLD. We examined recognition of happiness, sadness, fear and anger in facial expressions, non-verbal vocalisations and music in patients with FTLD and in healthy age matched controls. The FTLD group was significantly impaired in all modalities compared with controls, and this effect was most marked for music. Analysing each emotion separately, recognition of negative emotions was impaired in all three modalities in FTLD, and this effect was most marked for fear and anger. Recognition of happiness was deficient only with music. Our findings support the idea that FTLD causes impaired recognition of emotions across input channels, consistent with a common central representation of emotion concepts. Music may be a sensitive probe of emotional deficits in FTLD, perhaps because it requires a more abstract representation of emotion than do animate stimuli such as faces and voices.

Abstract

-* indicates shared first authorship -
The present study investigated the effects of selective attention on the processing of morphosyntactic errors in unattended parts of speech. Two groups of German native (L1) speakers participated in the present study. Participants listened to sentences in which irregular verbs were manipulated in three different conditions (correct, incorrect but attested ablaut pattern, incorrect and crosslinguistically unattested ablaut pattern). In order to track fast dynamic neural reactions to the stimuli, electroencephalography was used. After each sentence, participants in Experiment 1 performed a semantic judgement task, which deliberately distracted the participants from the syntactic manipulations and directed their attention to the semantic content of the sentence. In Experiment 2, participants carried out a syntactic judgement task, which put their attention on the critical stimuli. The use of two different attentional tasks allowed for investigating the impact of selective attention on speech processing and whether morphosyntactic processing steps are performed automatically. In Experiment 2, the incorrect attested condition elicited a larger N400 component compared to the correct condition, whereas in Experiment 1 no differences between conditions were found. These results suggest that the processing of morphosyntactic violations in irregular verbs is not entirely automatic but seems to be strongly affected by selective attention.

Abstract

This dissertation investigates the expression of spatial relationships in German Sign Language (Deutsche Gebärdensprache, DGS). The analysis focuses on linguistic expression in the spatial domain in two types of discourse: static scene description (location) and event narratives (location and motion). Its primary theoretical objectives are to characterize the structure of locative descriptions in DGS; to explain the use of frames of reference and perspective in the expression of location and motion; to clarify the interrelationship between the systems of frames of reference, signing perspective, and classifier predicates; and to characterize the interplay between iconicity principles, on the one hand, and grammatical and discourse constraints, on the other hand, in the use of these spatial devices. In more general terms, the dissertation provides a usage-based account of iconic mapping in the visual-spatial modality. The use of space in sign language expression is widely assumed to be guided by iconic principles, which are furthermore assumed to hold in the same way across sign languages. Thus, there has been little expectation of variation between sign languages in the spatial domain in the use of spatial devices. Consequently, perhaps, there has been little systematic investigation of linguistic expression in the spatial domain in individual sign languages, and less investigation of spatial language in extended signed discourse. This dissertation provides an investigation of spatial expressions in DGS by investigating the impact of different constraints on iconicity in sign language structure. The analyses have important implications for our understanding of the role of iconicity in the visual-spatial modality, the possible language-specific variation within the spatial domain in the visual-spatial modality, the structure of spatial language in both natural language modalities, and the relationship between spatial language and cognition

Abstract

Cognitive neuroscience approaches the brain as a cognitive system: a system that functionally is conceptualized in terms of information processing. We outline some aspects of this concept and consider a physical system to be an information processing device when a subclass of its physical states can be viewed as representational/cognitive and transitions between these can be conceptualized as a process operating on these states by implementing operations on the corresponding representational structures. We identify a generic and fundamental problem in cognition: sequentially organized structured processing. Structured sequence processing provides the brain, in an essential sense, with its processing logic. In an approach addressing this problem, we illustrate how to integrate levels of analysis within a framework of adaptive dynamical systems. We note that the dynamical system framework lends itself to a description of asynchronous event-driven devices, which is likely to be important in cognition because the brain appears to be an asynchronous processing system. We use the human language faculty and natural language processing as a concrete example through out.

Abstract

This dissertation investigates the roles of several probabilistic variables in the production and comprehension of reduced Dutch affixes. The central hypothesis is that linguistic units with a high probability of occurrence are
more likely to be reduced (Jurafsky et al., 2001; Aylett & Turk, 2004). This hypothesis is tested by analyzing the acoustic realizations of affixes, which are meaning-carrying elements embedded in larger lexical units. Most of the results prove to be compatible with the main hypothesis, but some appear to run counter to its predictions. The final chapter of the thesis discusses the implications of these findings for models of speech production, models of speech perception, and probability-based accounts of reduction.

Abstract

Two eye-tracking experiments tested whether native listeners can adapt
to reductions in casual Dutch speech. Listeners were exposed to segmental
([b] > [m]), syllabic (full-vowel-deletion), or no reductions. In a subsequent
test phase, all three listener groups were tested on how efficiently they could
recognize both types of reduced words. In the first Experiment’s exposure
phase, the (un)reduced target words were predictable. The segmental reductions
were completely consistent (i.e., involved the same input sequences).
Learning about them was found to be pattern-specific and generalized in the
test phase to new reduced /b/-words. The syllabic reductions were not consistent
(i.e., involved variable input sequences). Learning about them was
weak and not pattern-specific. Experiment 2 examined effects of word repetition
and predictability. The (un-)reduced test words appeared in the exposure
phase and were not predictable. There was no evidence of learning for
the segmental reductions, probably because they were not predictable during
exposure. But there was word-specific learning for the vowel-deleted words.
The results suggest that learning about reductions is pattern-specific and
generalizes to new words if the input is consistent and predictable. With
variable input, there is more likely to be adaptation to a general speaking
style and word-specific learning.

Abstract

Cognates and interlingual homographs are words that exist in multiple languages. Cognates, like “wolf” in Dutch and English, also carry the same meaning. Interlingual homographs do not: the word “angel” in English refers to a spiritual being, but in Dutch to the sting of a bee. The six experiments included in this thesis examined how these words are represented in the bilingual mental lexicon. Experiment 1 and 2 investigated the issue of task effects on the processing of cognates. Bilinguals often process cognates more quickly than single-language control words (like “carrot”, which exists in English but not Dutch). These experiments showed that the size of this cognate facilitation effect depends on the other types of stimuli included in the task. These task effects were most likely due to response competition, indicating that cognates are subject to processes of facilitation and inhibition both within the lexicon and at the level of decision making. Experiment 3 and 4 examined whether seeing a cognate or interlingual homograph in one’s native language affects subsequent processing in one’s second language. This method was used to determine whether non-identical cognates share a form representation. These experiments were inconclusive: they revealed no effect of cross-lingual long-term priming. Most likely this was because a lexical decision task was used to probe an effect that is largely semantic in nature. Given these caveats to using lexical decision tasks, two final experiments used a semantic relatedness task instead. Both experiments revealed evidence for an interlingual homograph inhibition effect but no cognate facilitation effect. Furthermore, the second experiment found evidence for a small effect of cross-lingual long-term priming. After comparing these findings to the monolingual literature on semantic ambiguity resolution, this thesis concludes that it is necessary to explore the viability of a distributed connectionist account of the bilingual mental lexicon.

Abstract

Spontaneously occurring speech is often seamlessly accompanied by hand gestures. Detailed
observations of video data suggest that speech and gesture are tightly synchronized in time,
consistent with a dynamic interplay between body and mind. However, spontaneous gesturespeech
synchrony has rarely been objectively quantified beyond analyses of video data, which
do not allow for identification of kinematic properties of gestures. Consequently, the point in
gesture which is held to couple with speech, the so-called moment of “maximum effort”, has
been variably equated with the peak velocity, peak acceleration, peak deceleration, or the onset
of the gesture. In the current exploratory report, we provide novel evidence from motiontracking
and acoustic data that peak velocity is closely aligned, and shortly leads, the peak pitch
(F0) of speech

Abstract

Cognitive research is often focused on experimental condition-driven reactions. Ethological studies frequently
rely on the observation of naturally occurring specific behaviors. In both cases, subjects are filmed during the
study, so that afterwards behaviors can be coded on video. Coding should typically be blind to experimental
conditions, but often requires more information than that present on video. We introduce a method for blindcoding
of behavioral videos that takes care of both issues via three main innovations. First, of particular
significance for playback studies, it allows creation of a “soundtrack” of the study, that is, a track composed of
synthesized sounds representing different aspects of the experimental conditions, or other events, over time.
Second, it facilitates coding behavior using this audio track, together with the possibly muted original video.
This enables coding blindly to conditions as required, but not ignoring other relevant events. Third, our method
makes use of freely available, multi-platform software, including scripts we developed.

Abstract

Dutch listeners' looks to printed words were tracked while they listened to instructions to click with their mouse on one of them. When presented with targets from word pairs where the first two syllables were segmentally identical but differed in stress location, listeners used stress information to recognize the target before segmental information disambiguated the words. Furthermore, the amount of lexical competition was influenced by the presence or absence of word-initial stress.

Abstract

Segmental as well as suprasegmental information is used by Dutch listeners to recognize words. The time-course of the effect of suprasegmental stress information on spoken-word recognition was investigated in a previous study, in which we tracked Dutch listeners' looks to arrays of four printed words as they listened to spoken sentences. Each target was displayed along with a competitor that did not differ segmentally in its first two syllables but differed in stress placement (e.g., 'CENtimeter' and 'sentiMENT'). The listeners' eye-movements showed that stress information is used to recognize the target before distinct segmental information is available. Here, we examine the role of durational information in this effect. Two experiments showed that initial-syllable duration, as a cue to lexical stress, is not interpreted dependent on the speaking rate of the preceding carrier sentence. This still held when other stress cues like pitch and amplitude were removed. Rather, the speaking rate of the preceding carrier affected the speed of word recognition globally, even though the rate of the target itself was not altered. Stress information modulated lexical competition, but did so independently of the rate of the preceding carrier, even if duration was the only stress cue present.

Abstract

This paper reports on the development of a flexible web based lexicon tool, LEXUS. LEXUS is targeted at linguists involved in language documentation (of endangered languages). It allows the creation of lexica within the structure of the proposed ISO LMF standard and uses the proposed concept naming conventions from the ISO data categories, thus enabling interoperability, search and merging. LEXUS also offers the possibility to visualize language, since it provides functionalities to include audio, video and still images to the lexicon. With LEXUS it is possible to create semantic network knowledge bases, using typed relations. The LEXUS tool is free for use. Index Terms: lexicon, web based application, endangered languages, language documentation.

Abstract

Thematic roles such as Agent and Instrument have a long-standing place in theories of event representation. Nonetheless, the structure of these categories has been difficult to determine. We investigated how instrumental events, such as someone slicing bread with a knife, are categorized in English. Speakers described a variety of typical and atypical instrumental events, and we determined the similarity structure of their descriptions using correspondence analysis. We found that events where the instrument is an extension of an intentional agent were most likely to elicit similar language, highlighting the importance of agency in structuring instrumental categories.

Abstract

The recognition of emotions of disgust, anger and fear have been shown to be significantly impaired in Huntington’s disease (eg,Sprengelmeyer et al, 1997, 2006; Gray et al, 1997; Milders et al, 2003,Montagne et al, 2006; Johnson et al, 2007; De Gelder et al, 2008). The relative impairment of these emotions might have implied a recognition impairment specific to negative emotions. Could the asymmetric recognition deficits be due not to the complexity of the emotion but rather reflect the complexity of the task? In the current study, 15 Huntington’s patients and 16 control subjects were presented with negative and positive non-speech emotional vocalisations that were to be identified as anger, fear, sadness, disgust, achievement, pleasure and amusement in a forced-choice paradigm. This experiment more accurately matched the negative emotions with positive emotions in a homogeneous modality. The resulting dually impaired ability of Huntington’s patients to identify negative and positive non-speech emotional vocalisations correctly provides evidence for an overall emotional recognition deficit in the disease. These results indicate that previous findings of a specificity in emotional recognition deficits might instead be due to the limitations of the visual modality. Previous experiments may have found an effect of emotional specificy due to the presence of a single positive emotion, happiness, in the midst of multiple negative emotions. In contrast with the previous literature, the study presented here points to a global deficit in the recognition of emotional sounds.

Abstract

Wat doen onze ogen als we met andere mensen praten? In zijn proefschrift beschrijft Federico Rossano hoe mensen hun ogen gebruiken tijdens face-to-face interacties. Onze oogbewegingen blijken opvallend geordend en voorspelbaar: zo is het bijvoorbeeld mogelijk om met uitsluitend de ogen een reactie uit te lokken als de gesprekspartner niet direct reageert. Ook wanneer bijvoorbeeld een vraag-antwoordreeks ten einde loopt, coördineren gespreksdeelnemers hun oogbewegingen op een specifieke manier. Daarnaast heeft luisteren naar een verhaal of luisteren naar een vraag verschillende implicaties voor oogbewegingen. Dit proefschrift bevat daarom belangrijke informatie voor experts op het gebied van kunstmatige intelligentie en computerwetenschappers: de voorspelbaarheid en reproduceerbaarheid van natuurlijke oogbewegingen kan onder andere gebruikt worden bij de ontwikkeling van robots of avatars.

Abstract

In this paper, I will give an overview of some well-studied multimodal signals that humans produce while they communicate with other humans, and discuss the implications of those studies for HCI. I will first discuss a conceptual framework that allows us to distinguish between functional and sensory modalities. This distinction is important, as there are multiple functional modalities using the same sensory modality (e.g., facial expression and eye-gaze in the visual modality). A second theoretically important issue is redundancy. Some signals appear to be redundant with a signal in another modality, whereas others give new information or even appear to give conflicting information (see e.g., the work of Susan Goldin-Meadows on speech accompanying gestures). I will argue that multimodal signals are never truly redundant. First, many gestures that appear at first sight to express the same meaning as the accompanying speech generally provide extra (analog) information about manner, path, etc. Second, the simple fact that the same information is expressed in more than one modality is itself a communicative signal. Armed with this conceptual background, I will then proceed to give an overview of some multimodalsignals that have been investigated in human-human research, and the level of understanding we have of the meaning of those signals. The latter issue is especially important for potential implementations of these signals in artificial agents. First, I will discuss pointing gestures. I will address the issue of the timing of pointing gestures relative to the speech it is supposed to support, the mutual dependency between pointing gestures and speech, and discuss the existence of alternative ways of pointing from other cultures. The most frequent form of pointing that does not involve the index finger is a cultural practice called lip-pointing which employs two visual functional modalities, mouth-shape and eye-gaze, simultaneously for pointing. Next, I will address the issue of eye-gaze. A classical study by Kendon (1967) claims that there is a systematic relationship between eye-gaze (at the interlocutor) and turn-taking states. Research at our institute has shown that this relationship is weaker than has often been assumed. If the dialogue setting contains a visible object that is relevant to the dialogue (e.g., a map), the rate of eye-gaze-at-other drops dramatically and its relationship to turn taking disappears completely. The implications for machine generated eye-gaze are discussed. Finally, I will explore a theoretical debate regarding spontaneous gestures. It has often been claimed that the class of gestures that is called iconic by McNeill (1992) are a “window into the mind”. That is, they are claimed to give the researcher (or even the interlocutor) a direct view into the speaker’s thought, without being obscured by the complex transformation that take place when transforming a thought into a verbal utterance. I will argue that this is an illusion. Gestures can be shown to be specifically designed such that the listener can be expected to interpret them. Although the transformations carried out to express a thought in gesture are indeed (partly) different from the corresponding transformations for speech, they are a) complex, and b) severely understudied. This obviously has consequences both for the gesture research agenda, and for the generation of iconic gestures by machines.

Abstract

This paper presents the first application of polynomial modeling as a means for validating phonological pitch accent labels to German data. It is compared to traditional phonetic analysis (measuring minima, maxima, alignment). The traditional method fares better in classification, but results are comparable in statistical accent pair testing. Robustness tests show that pitch correction is necessary in both cases. The approaches are discussed in terms of their practicability, applicability to other domains of research and interpretability of their results.

Abstract

In the context of the source-filter theory of speech, it is well established that intelligibility is heavily reliant on information carried by the filter, that is, spectral cues (e.g., Faulkner et al., 2001; Shannon et al., 1995). However, the extraction of other types of information in the speech signal, such as emotion and identity, is less well understood. In this study we investigated the extent to which emotion recognition in speech depends on filterdependent cues, using a forced-choice emotion identification task at ten levels of noise-vocoding ranging between one and 32 channels. In addition, participants performed a speech intelligibility task with the same stimuli. Our results indicate that compared to speech intelligibility, emotion recognition relies less on spectral information and more on cues typically signaled by source variations, such as voice pitch, voice quality, and intensity. We suggest that, while the reliance on spectral dynamics is likely a unique aspect of human speech, greater phylogenetic continuity across species may be found in the communication of affect in vocalizations.

Abstract

Research using event-related brain potentials (ERPs) has demonstrated an early differential effect in fronto-central regions when processing emotional, as compared to affectively neutral facial stimuli (e.g., Eimer & Holmes, 2002). In this talk, data demonstrating a similar effect in the auditory domain will be presented. ERPs were recorded in a one-back task where participants had to identify immediate repetitions of emotion category, such as a fearful sound followed by another fearful sound. The stimulus set consisted of non-verbal emotional vocalisations communicating positive and negative sounds, as well as neutral baseline conditions. Similarly to the facial domain, fear sounds as compared to acoustically controlled neutral sounds, elicited a frontally distributed positivity with an onset latency of about 150 ms after stimulus onset. These data suggest the existence of a rapid multi-modal frontocentral mechanism discriminating emotional from non-emotional human signals.

Abstract

The difficulty of the task of segmenting a speech signal into its words is immediately clear when listening to a foreign language; it is much harder to segment the signal into its words, since the words of the language are unknown. Infants are faced with the same task when learning their first language. This study provides a better understanding of the task that infants face while learning their native language. We employed an automatic algorithm on the task of speech segmentation without prior knowledge of the labels of the phonemes. An analysis of the boundaries erroneously placed inside a phoneme showed that the algorithm consistently placed additional boundaries in phonemes in which acoustic changes occur. These acoustic changes may be as great as the transition from the closure to the burst of a plosive or as subtle as the formant transitions in low or back vowels. Moreover, we found that glottal vibration may attenuate the relevance of acoustic changes within obstruents. An interesting question for further research is how infants learn to overcome the natural tendency to segment these ‘dynamic’ phonemes.

Abstract

Articulatory feature (AF) modelling of speech has received a considerable amount of attention in automatic speech recognition research. Although termed ‘articulatory’, previous definitions make certain assumptions that are invalid, for instance, that articulators ‘hop’ from one fixed position to the next. In this paper, we studied two methods, based on support vector classification (SVC) and regression (SVR), in which the articulation continuum is modelled without being restricted to using discrete AF value classes. A comparison with a baseline system trained on quantised values of the articulation continuum showed that both SVC and SVR outperform the baseline for two of the three investigated AFs, with improvements up to 5.6% absolute.

Abstract

Listeners outperform ASR systems in every speech recognition task. However, what is not clear is where this human advantage originates. This paper investigates the role of acoustic feature representations. We test four (MFCCs, PLPs, Mel Filterbanks, Rate Maps) acoustic representations, with and without ‘pitch’ information, using the same backend. The results are compared with listener results at the level of articulatory feature classification. While no acoustic feature representation reached the levels of human performance, both MFCCs and Rate maps achieved good scores, with Rate maps nearing human performance on the classification of voicing. Comparing the results on the most difficult articulatory features to classify showed similarities between the humans and the SVMs: e.g., ‘dental’ was by far the least well identified by both groups. Overall, adding pitch information seemed to hamper classification performance.

Abstract

In foreign-accented speech, pronunciation typically deviates from the canonical form to some degree. For native listeners, it has been shown that word recognition is more difficult for strongly-accented words than for less strongly-accented words. Furthermore recognition of strongly-accented words becomes easier with additional exposure to the foreign accent. In this paper, listeners’ behaviour was simulated with Fine-tracker, a computational model of word recognition that uses real speech as input. The simulations showed that, in line with human listeners, 1) Fine-Tracker’s recognition outcome is modulated by the degree of accentedness and 2) it improves slightly after brief exposure with the accent. On the level of individual words, however, Fine-tracker failed to correctly simulate listeners’ behaviour, possibly due to differences in overall familiarity with the chosen accent (German-accented Dutch) between human listeners and Fine-Tracker.

Abstract

There is now considerable evidence that fine-grained acoustic-phonetic detail in the speech signal helps listeners to segment a speech signal into syllables and words. In this paper, we compare two computational models of word recognition on their ability to capture and use this finephonetic detail during speech recognition. One model, SpeM, is phoneme-based, whereas the other, newly developed Fine- Tracker, is based on articulatory features. Simulations dealt with modelling the ability of listeners to distinguish short words (e.g., ‘ham’) from the longer words in which they are embedded (e.g., ‘hamster’). The simulations with Fine- Tracker showed that it was, like human listeners, able to distinguish between short words from the longer words in which they are embedded. This suggests that it is possible to extract this fine-phonetic detail from the speech signal and use it during word recognition.

Abstract

Aging often affects sensitivity to the higher frequencies, which results in the loss of sensitivity to phonetic detail in speech. Hearing loss may therefore interfere with the categorisation of two consonants that have most information to differentiate between them in those higher frequencies and less in the lower frequencies, e.g., /f/ and /s/. We investigate two acoustic cues, i.e., formant transitions and fricative intensity, that older listeners might use to differentiate between /f/ and /s/. The results of two phonetic categorisation tasks on 38 older listeners (aged 60+) with varying degrees of hearing loss indicate that older listeners seem to use formant transitions as a cue to distinguish /s/ from /f/. Moreover, this ability is not impacted by hearing loss. On the other hand, listeners with increased hearing loss seem to rely more on intensity for fricative identification. Thus, progressive hearing loss may lead to gradual changes in perceptual cue weighting.

Abstract

Young listeners can quickly modify their interpretation of a speech sound when a talker produces the sound ambiguously. Young Dutch listeners rely mainly on the higher frequencies to distinguish between /f/ and /s/, but these higher frequencies are particularly vulnerable to age-related hearing loss. We therefore tested whether older Dutch listeners can show perceptual retuning given an ambiguous pronunciation in between /f/ and /s/. Results of a lexically-guided perceptual learning experiment showed that older Dutch listeners are still able to learn non-standard pronunciations of /f/ and /s/. Possibly, the older listeners have learned to rely on other acoustic cues, such as formant transitions, to distinguish between /f/ and /s/. However, the size and duration of the perceptual effect is influenced by hearing loss, with listeners with poorer hearing showing a smaller and a shorter-lived learning effect.