Unfolding Prosody Guides the Development of Word Segmentation

Chinese automatic word segmentation is the premise of Chinese information processing, which is widely used in Chinese full-text retrieval, Chinese automatic full-text translation, Chinese text-to-speech conversion (TTS) and other fields. A dictionary plays an important role in Chinese word segmentation. The advantages and disadvantages of the word segmentation mechanism directly affect the speed and efficiency of Chinese word segmentation. Therefore, we propose a deep learning method for Chinese word segmentation. First, a separable convolution bidirectional long and short-term memory condition random field word segmentation model with feature points containing dictionary features is constructed. The model parameters are obtained by training on the existing word segmentation corpus. Then, the software engineering field text is used as the small-scale word segmentation training corpus. The word segmentation model of general corpus is fine-tuned. The experimental results show that the transfer learning reduces the iteration times of the domain segmentation model. Meanwhile, compared with other Chinese word segmentation models, the proposed model can reduce the corpus labeling time in training process and realize the cross-domain transfer learning of word segmentation model.

Event Abstract Back to Event Statistical segmentation of streams of syllables: a pilot EEG study Leidy J. Castro-Meneses1*, Paul Sowman1 and Blake W. Johnson1 1 Macquarie University, Department of Cognitive Science, Australia Previous studies have claimed that before infants learn prosodic patterns (e.g., stress, pitch) our brain uses statistical mechanisms that support word segmentation by tracing the transitional probabilities of syllables. We were interested in exploring whether word extraction is affected when word components are separated by silence gaps. Thus, although the word syllables are adjacent, they may be processed individually and not grouping into a word-like unit. It was hypothesised that word segmentation would be highly disrupted when silence gaps are either longer than the syllable length or have unsystematic lengths. We record electrophysiological brain signals from 8 Native English speakers while they learned 8 artificial languages presented randomly. Each artificial language consisted of nine trisyllabic pseudowords (made up of consonant + vowel= cv + cv + cv). After that, learning of each language was assessed in a two alternative forced choice paradigm. The analysis of the event-related potentials (ERPs) showed the N400 component, which was enhanced for word onsets in those languages that were better recognised behaviourally. The behavioural analysis showed that participants were able to learn the words of those artificial languages when the silence gaps were shorter than a syllable length. When the gaps where longer or mixed, the responses do not differ from chance. These findings show that the brain is able to learn/group temporary non-adjacent elements. A useful learning mechanism when infants are learning for example phrases. However, further research will be required to extend these results. Acknowledgements This research was funded by the Australian Research Council Centre of Excellence in Cognition and its Disorders (CE110001021) http://www.ccd.edu.au. References Buiatti, M., Peña, M., and Dehaene-Lambertz, G. (2008). Investigating the neural correlates of continuous speech computation with frequency-tagged neuroelectric responses. NeuroImage, 44(2), 509–519. DOI: 10.1016/j.neuroimage.2008.09.015 Creel, S., Newport, E., and Aslin, R. (2004). Distand melodies: Statistical learning of nonadjacent dependencies in tone sequences. Journal of Experimental Psychology: Learning, memory and cognition, 30(5), 1119-1130. DOI: 10.1037/0278-7393.30.5.1119 De Diego Balaguer, R., Toro, J.M., Rodriguez-Fornells, A., and Bachoud-Lévi, A.C. (2007). Different Neurophysiological Mechanisms Underlying Word and Rule Extraction from Speech. PLoS ONE, 2(11): e1175. DOI:10.1371/journal.pone.0001175 Endress, A., and Mehler, J. (2009). The surprising power of statistical learning: When fragment knowledge leads to false memories of unheard words. Journal of Memory and language, 60, 351-367. DOI: 10.1016/j.jml.2008.10.003 Gebhart, A., Newport, E., and Aslin, R. (2009). Statistical learning of adjacent and nonadjacent dependencies among nonlinguistic sounds. Psychonomic Bulletin & Review, 16 (3), 486-490. DOI: 10.3758/PBR.16.3.486 Luo, H., and Poeppel, D. (2012). Cortical oscillations in auditory perception and speech: evidence for two temporal windows in human auditory cortex. Frontiers Psychology, 3(170). DOI: 10.3389/fpsyg.2012.00170 Marcus, G.F., Vijayan, S., Rao, B., and Vishton, M. (1999). Rule learning by seven-month-old infants. Science, 283 (5398), 77-80. DOI: 10.1126/science.283.5398.77 Newport, E., and Aslin, R. (2004). Learning at a distance I. Statistical learning of non-adjacent dependencies. Cognitive Psychology, 48(2), 127-162. DOI: 10.1016/S0010-0285(03)00128-2 Pelucchi, B., Hay, J., and Saffran, J. (2009). Statistical learning in a natural language by 8-Month-old infants. Journal of Child development, 80(3), 674-685. DOI: 10.1111/j.1467-8624.2009.01290.x Peña, M., Bonatti, L., Nespor, M., and Mehler, J. (2002). Signal-driven computations in speech procession. Science, 298 (5593), 604-607. DOI: 10.1126/science.1072901 Romberg, A., and Saffran, J. (2010). Statistical learning and language acquisition. Wiley Interdisciplinary reviews: Cognitive Science, 1 (6), 906-914. DOI: 10.1002/wcs.78 Saffran, J.R., Aslin, R.N., and Newport, E.L (1996). Statistical learning by 8-month-old infants. Science, 274, 1926-1928 Saffran, J.R., Newport, E.L., and Aslin, R.N (1996). word segmentation: the role of distributional cues. journal of Memory and Language, 35, 606-621. Teinonen, T., and Huotilaine, M. (2011). Implicit Segmentation of a Stream of Syllables based on transitional probabilities: an MEG study. Journal of Psycholinguist Research, 14, 71-82. DOI: 10.1007/s10936-011-9182-2 Keywords: statistical learning, word segmentation, artificial language, EEG, Transitional probabilities Conference: ACNS-2012 Australasian Cognitive Neuroscience Conference, Brisbane, Australia, 29 Nov - 2 Dec, 2012. Presentation Type: Poster Presentation Topic: Language Citation: Castro-Meneses LJ, Sowman P and Johnson BW (2012). Statistical segmentation of streams of syllables: a pilot EEG study. Conference Abstract: ACNS-2012 Australasian Cognitive Neuroscience Conference. doi: 10.3389/conf.fnhum.2012.208.00129 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 13 Oct 2012; Published Online: 17 Nov 2012. * Correspondence: Mrs. Leidy J Castro-Meneses, Macquarie University, Department of Cognitive Science, Sydney, Australia, l.castro-meneses@westernsydney.edu.au Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Leidy J Castro-Meneses Paul Sowman Blake W Johnson Google Leidy J Castro-Meneses Paul Sowman Blake W Johnson Google Scholar Leidy J Castro-Meneses Paul Sowman Blake W Johnson PubMed Leidy J Castro-Meneses Paul Sowman Blake W Johnson Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Word segmentation, detecting word boundaries in continuous speech, is a critical aspect of language learning. Previous research in infants and adults demonstrated that a stream of speech can be readily segmented based solely on the statistical and speech cues afforded by the input. Using functional magnetic resonance imaging (fMRI), the neural substrate of word segmentation was examined on-line as participants listened to three streams of concatenated syllables, containing either statistical regularities alone, statistical regularities and speech cues, or no cues. Despite the participants' inability to explicitly detect differences between the speech streams, neural activity differed significantly across conditions, with left-lateralized signal increases in temporal cortices observed only when participants listened to streams containing statistical regularities, particularly the stream containing speech cues. In a second fMRI study, designed to verify that word segmentation had implicitly taken place, participants listened to trisyllabic combinations that occurred with different frequencies in the streams of speech they just heard ("words," 45 times; "partwords," 15 times; "nonwords," once). Reliably greater activity in left inferior and middle frontal gyri was observed when comparing words with partwords and, to a lesser extent, when comparing partwords with nonwords. Activity in these regions, taken to index the implicit detection of word boundaries, was positively correlated with participants' rapid auditory processing skills. These findings provide a neural signature of on-line word segmentation in the mature brain and an initial model with which to study developmental changes in the neural architecture involved in processing speech cues during language learning.

Event-related potentials (ERPs) were recorded during a picture naming task of simple and complex words in children with typical speech and with childhood apraxia of speech (CAS). Results reveal reduced amplitude prior to speaking complex (multisyllabic) words relative to simple (monosyllabic) words for the CAS group over the right hemisphere during a time window thought to reflect phonological encoding of word forms. Group differences were also observed prior to production of spoken tokens regardless of word complexity during a time window just prior to speech onset (thought to reflect motor planning/programming). Results suggest differences in pre-speech neurolinguistic processes.

Effects of Morphological Complexity on Phonological Output Deficits in Fluent and Nonfluent Aphasia

This study explored the role of consonants and vowels in auditory word perception in Hebrew, using an auditory-auditory priming procedure. A common finding is the greater importance for consonants than vowels in spoken word processing, henceforth the C-bias. What underlies the C-bias is a controversial issue between the "acquired functional hypothesis" and the "initial-bias hypothesis." The former predicts a different pattern of asymmetry among languages, while the latter claims that the C-bias exists in all languages. Hebrew is interesting because it allows the exploration of whether language-specific morphological properties can modulate consonant/vowel asymmetry. This opportunity stems from its unique complex Semitic morphological structure that was found to have a central role in lexical access. To evaluate the consonant/vowel asymmetry in Hebrew in relation to previously studied languages, recognition of morphologically complex Hebrew words was compared with recognition of foreign Hebrew words that do not have the typical Hebrew morphological complex structure, hereafter simple words. The results showed a significantly stronger C-bias for morphologically complex words than for simple words. We suggest that this within-variation supports the acquired functional hypothesis, so that even within one language different patterns of consonant/vowel asymmetry are revealed depending on the internal morphological structure of the word. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Songs as an aid for language acquisition

The next step of isolated word recognition will be the sentence or continuous speech recognition. In this paper, an algorithm to estimates maximum posterior probabilities from input chhattisgarhi speech has proposed. To compute the accurate and precise maximum likelihoods, one need to extract the features from segmented words accurately. An implemented algorithm for automatic word segmentation gives up to 96.00% of average word boundary detection accuracy. Mel filter cepstral coefficients, delta cepstral and delta-delta cepstral speech characteristics have been extracted from the signals. Automatic segmented words from the speech have added and created a vocabulary. Further, 3-gram language model has implemented to resolve the ambiguity of similar pronunciation of the words. The last stage of the system is to compute maximum log likelihoods and maximum posterior probabilities of the word sequence. The experiments have been carried out on self-recorded 220 continuous chhattisgarhi speech, which consist of 2640 sentences and 330,000 words. We have used multilayer feedforward neural network (MLP) and Hidden markov model (HMM) for maximum likelihood estimation (MLE). The details of experimental analysis have presented in the paper.

A hierarchical approach to recognition of handwritten Bangla characters

Preface. How to Use this Book. Section A. Introduction. 1. The Study of Language. Introduction. Why Study Language and Why Is It So Difficult? What is Language? How Has Language Changed Over Time? What is Language For? The History and Methods of Psycholinguistics. Models in Psycholinguistics. Language and the Brain. Themes and Controversies. Summary. Some Questions to Think About. Further Reading. 2. Describing Language. Introduction. How to Describe Speech Sounds. Consonants. Vowels. Syllables. Linguistic Approaches to Syntax. Summary. Some Questions to Think About. Further Reading. Section B. The Biological and Developmental Bases of Language. 3. The Foundations of Language. Introduction. Where Did Language Come From? Do Animals Have Language? The Biological Basis of Language. Is There a Critical Period for Language Development? The Cognitive Basis of Language. The Social Basis of Language. What is the Relation Between Language and Thought? Summary. Some Questions to Think About. Further Reading. 4. Language Development. Introduction. What Drives Language Development. The Language Acquisition Device. Phonological Development. Syntactic Development. Summary. Some Questions to Think About. Further Reading. 5. Bilingualism and Second Language Acquisition. Introduction. Bilingualism. Second Language Acquisition. Evaluation of Work on Bilingualism and Second Language Acquisition. Summary. Some Questions to Think About. Further reading. Section C. Word Recognition. 6. Recognizing Visual Words. Introduction. Basic Methods and Findings. What Makes Word Recognition Easier (or Harder)? Attentional Processes in Visual Word Recognition. Do Different Tasks Give Consistent Results? Is There a Dedicated Visual Word Recognition System? Meaning-based Facilitation of Visual Word Recognition. Processing Morphologically Complex Words. Models of Visual Word Recognition. Coping with Lexical Ambiguity. Summary. Some Questions to Think About. Further Reading. 7. Reading. Introduction. The Writing System. A Preliminary Model of Reading. The Processes of Normal Reading. The Neuroscience of Adult Reading Disorders. Models of Word Naming. Connectionism Models of Dyslexia. Comparison of Models. Summary. Some Questions to Think About. Further Reading. 8. Learning to Read and Spell. Introduction. Normal Reading Development. Phonological Awareness. How Should Reading be Taught? Learning to Spell. Developmental Dyslexia. Summary. Some Questions to Think About. Further Reading. 9. Understanding Speech. Introduction. Recognizing Speech. Models of Speech Recognition. The Neuroscience of Spoken Word Recognition. Summary. Some Questions to Think About. F urther Reading. Section D. Meaning and Using Language. 10. Understanding the Structure of Sentences. Introduction. Dealing with Structural Ambiguity. Early work on Parsing. Processing Structural Ambiguity. Gaps, Traces, and Unbounded Dependencies. The Neuroscience of Parsing. Summary. Some Questions to Think About. Further Reading. 11. Word Meaning. Introduction. Classic Approaches to Semantics. Semantic Networks. Semantic Features. Family Resemblance Models. Combining Concepts. Figurative Language. The Neuroscience of Semantics. Summary. Some Questions to Think About. Further Reading. 12. Comprehension. Introduction. Memory for Text and Inferences. Inferences. References and Ambiguity. Models of Text Processing. Individual Differences in Comprehension Skills. The Neuroscience of Text and Discourse Processing. Summary. Some Questions to Think About. Further Reading. Section E. Production and Other Aspects of Language. 13. Language Production. Introduction. Slips of the Tongue. Syntactic Planning. Lexicalization. Phonological Encoding. The Analysis of Hesitations. The Neuroscience of Speech Production. Writing and Agraphia. Summary. Some Questions to Think About. Further Reading. 14. How Do We Use Language? Introduction. Making Inferences in Conversation. The Structure of Conversation. Collaboration in Dialog. Sound and Vision. Summary. Some Questions to Think About. Further Reading. 15. The Structure of the Language System. Introduction. What are the Modules of Language? How Many Lexicons are There? Language and Short-term Memory. Summary. Some Questions to Think About. Further Reading. 16. New Directions. Introduction. Themes in Psycholinguistics Revisited. Some Growth Areas? Conclusion. Appendix: Connectionism. Interactive Activation Models. Back-propagation. Further Reading. Glossary. Example of Sentence Analysis. References. Author Index. Subject Index.

How is a word visually recognized? We investigated word recognition by employing a novel psychophysical task in which word segments (true morphemes or nonmorphemic orthographic sequences) were colored blue/red, red/blue, or black, while participants wore anaglyph glasses. The color combinations split the words either at the morpheme boundary (legal split) or at one character to the left or to the right of the split (illegal split). This manipulation divided the retinal information along the fovea, allowing us to project the different word segments to both left and right hemispheres via contralateral and ipsilateral visual pathways. Our goal was to investigate the roles of morphology and semantics in the early moments of visual word recognition, potentially tapping functional properties of the visual word form area. Participants (N = 72) performed a masked lexical decision task (60 ms exposure) to semantically transparent compounds (e.g., football), monomorphemic pseudocompounds (e.g., shamrock), and unsegmentable monomorphemic words (e.g., jeopardy). We found an effect of legality for compounds and pseudocompounds, when presented via the stronger contralateral visual pathways, suggesting an early morpho-orthographic segmentation, but with true compounds yielding faster and more accurate responses than pseudocompounds across all projection combinations. Our results are compatible with a model of visual word recognition involving an initial stage of morpho-orthographic analysis that is insensitive to semantics, followed by later stages of morphological computations and semantic access. Although our evidence relies on a psychophysical task, we suggest that these processing stages are in line with a functional division between visual word form area-1 (orthographic) and visual word form area-2 (lexical) in the early visual word recognition system. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

A series of eye-tracking and categorization experiments investigated the use of speaking-rate information in the segmentation of Dutch ambiguous-word sequences. Juncture phonemes with ambiguous durations (e.g., [s] in 'eens (s)peer,' "once (s)pear," [t] in 'nooit (t)rap,' "never staircase/quick") were perceived as longer and hence more often as word-initial when following a fast than a slow context sentence. Listeners used speaking-rate information as soon as it became available. Rate information from a context proximal to the juncture phoneme and from a more distal context was used during on-line word recognition, as reflected in listeners' eye movements. Stronger effects of distal context, however, were observed in the categorization task, which measures the off-line results of the word-recognition process. In categorization, the amount of rate context had the greatest influence on the use of rate information, but in eye tracking, the rate information's proximal location was the most important. These findings constrain accounts of how speaking rate modulates the interpretation of durational cues during word recognition by suggesting that rate estimates are used to evaluate upcoming phonetic information continuously during prelexical speech processing.

Methods for text segmentation from scene images

Multiple Sclerosis (MS) is a neurological disease characterised by damage affecting large bundles of white matter fibres. Morphological segmentation of complex words (e.g. walked) into stems (walk) and suffixes (∼ed) is thought to depend on intact white matter. We tested the hypothesis that Arabic speaking patients with MS may lose the ability to segment morphologically complex words in a primed lexical decision task using word pairs that shared either a root and a semantic relationship (+R + S, e.g. “AnzAl”–“nuzwl” lowering-landing), a root without semantics (+R–S, e.g. “rtAbp”–“trtyb” monotony-tidying up),a semantic relationship (–R + S, e.g. “xyr”–“nEmp” good-grace), or a phonological relationship (–R + Phon, e.g. “mEdn”–“mEAnd” mineral-stubborn). While healthy controls showed priming by root regardless of semantics and inhibition by phonology, the patients showed facilitation by semantics (+R + S and –R + S), and inhibition by phonology (–R + Phon). These findings are used to adjudicate three contending models of lexical processing.

Word segmentation in documents is a critical stage towards word and character recognition, as well as word spotting. Despite recent advancements in word segmentation and object detection, detecting instances of words in a cluttered handwritten document remains a non-trivial task that requires a large amount of labeled documents for training. We present a flexible and general framework for word segmentation in handwritten documents, which incorporates techniques from the recent object detection literature as well as document analysis tools. Our method utilizes information that is relevant for word segmentation and ignores other highly variable information contained in a handwritten text, thus allowing for efficient transfer learning between datasets and alleviating the need for labeled training data. Our approach efficiently detects words in a variety of scanned document images, including historical handwritten documents and modern day handwritten documents, presenting excellent results on existing benchmarks. In addition, we demonstrate the usefulness of our approach by achieving state-of-the-art results for segmentation-free word spotting tasks.