Human Speech Perception 1
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Heads And Tails in Word Perception: Evidence For `Early-to-Late' Processing in Listening and Reading
Authors:
Sieb G. Nooteboom, Utrecht Institute of Linguistics (The Netherlands)
Meinou van Dijk, Utrecht University (The Netherlands)
Page (NA) Paper number 117
Abstract:
Sequential models of word perception assigned a very special role to word onsets. This accounted in a natural way for evidence that lexical access is easier from word beginnings than from word endings, a property speech perception shares with reading. Sequential models badly failed on other scores, however. More recent competition models seem to give equal weight to stimulus information, independent of position within the word. The present word recognition experiment aimed at testing the hypothesis that, other things being equal, mismatches are more damaging to word perception at onsets than at offsets of embedded words, both in speech perception and in reading. Results show that word recognition is quite good in all conditions, even when word onsets are mutilated, and mis-timed, thus lending support to competition models. Yet, the results also show that lexical access is modulated by some early-to-late or left-to-right component, as if human word perception displays a mixture of sequential and competition processing.
SL980117.PDF (From Author)
SL980117.PDF (Rasterized)
TOP
Evidence for Early Effects of Sentence Context on Word Segmentation
Authors:
Saskia te Riele, Utrecht University, UiL-OTS (The Netherlands)
Hugo Quené, Utrecht University, UiL-OTS (The Netherlands)
Page (NA) Paper number 326
Abstract:
The present paper focuses on the segmentation of two-word phrases containing two closely competing lexical hypotheses. It is hypothesized that the bottom-up information, which also includes a mechanism called the Possible-Word Constraint, is explored first in segmenting these phrases. Non-sensory sentential information influences this process at a later stage and only shows an effect if the bottom-up information does not lead to one dominating interpretation. The results of the present experiment show that beside the acoustic information listeners can and do use contextual information at a relatively early moment, at which the two possible segmentations are both still active and the bottom-up information has not yet suppressed the acoustically inconsistent interpretation. This effect became apparent, since disambiguating bottom-up information arrived relatively late in the stimulus phrases. Hence, it was concluded that both sensory and non-sensory information are employed to affect activation levels of competing lexical hypotheses at an early moment.
SL980326.PDF (From Author)
SL980326.PDF (Rasterized)
TOP
Assimilation and Anticipation in Word Perception
Authors:
Hugo Quené, Utrecht institute of Linguistics OTS, Utrecht University (The Netherlands)
Maya van Rossum, Utrecht institute of Linguistics OTS, Utrecht University (The Netherlands)
Mieke van Wijck, Utrecht institute of Linguistics OTS, Utrecht University (The Netherlands)
Page (NA) Paper number 113
Abstract:
Words in connected speech are often assimilated to subsequent words. Some property of that upcoming word may then be determined in advance; these advance assimilatory cues may facilitate perception of that word. A gating experiment was conducted in Dutch, studying anticipatory voice assimilation between plosives, in 24 two-word combinations. In Dutch, voicing in a word-final plosive can only be caused by anticipatory assimilation to the next, voiced initial plosive, e.g. "rie[db]lint". Voiced and unvoiced variants of final and initial plosives were cross-spliced. Responses for assimilated, voiced-final stimuli show a strong bias to voiced-initial responses, as predicted. Even at longer gates in the hybrid condition "rie[dp]lint", after hearing the unvoiced initial plosive, listeners often came up with a voiced-initial response, with high confidence. Hence, advance phonological 'voiced-initial' cues were often stronger than acoustic 'unvoiced-initial' cues. These gating results suggest that listeners use advance assimilatory cues in word perception.
SL980113.PDF (From Author)
SL980113.PDF (Rasterized)
TOP
Lexical Activation by Assimilated and Reduced Tokens
Authors:
M. Louise Kelly, University of Edinburgh (U.K.)
Ellen Gurman Bard, University of Edinburgh (U.K.)
Catherine Sotillo, University of Edinburgh (U.K.)
Page (NA) Paper number 565
Abstract:
Running speech contains abundant assimilated and phonologically reduced tokens, but there is considerable debate about how such varied pronunciations disrupt access to the corresponding words in the listeners' mental lexicons. While previous studies have examined the effects of carefully produced or electronically edited reductions, we present two experiments which compare cross-modal repetition priming for lexical decisions by more reduced spontaneous forms and less reduced read forms of the same words uttered by the same speakers in the same phrases. Though less priming is found for the more reduced spontaneous tokens, both versions of words produce significant priming effects, whether the majority of stimuli are taken from spontaneous speech (Experiment 1) or from read speech (Experiment 2). Priming is more robust if tokens themselves contain the context licensing reduction.