Comparison of Syllabification Algorithms and Training Strategies for Robust Word Count Estimation across Different Languages and Recording Conditions

Räsänen, Okko; Seshadri, Shreyas; Casillas, Marisa

doi:10.21437/interspeech.2018-1047

Cited by 9 publications

(14 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nevertheless, this downside of "traditional" syllabification algorithms can potentially be tackled with machine learning approaches. By training the system using a large number of syllables from different languages (potentially from different noise conditions), a machine learning algorithm may learn a general solution for syllable counting that tolerates nonspeech audio content better than the traditional signal processing approaches (e.g., Räsänen, Seshadri, & Casillas, 2018b;. This makes syllables a potential candidate unit that could be estimated directly from speech instead of using correlated features such as speech duration or phone counts as linear correlates of syllable counts.…”

Section: Technical Considerationsmentioning

confidence: 99%

“…ALICE uses the SylNet baseline model from the original paper that is trained on approximately 10 hours of handannotated Estonian and Korean speech, but further adapting the model to the present daylong child-centered data using the standard adaptation procedure of the algorithm (see Experimental Setup for details). In our initial experiments on child-centered audio, SylNet was also compared to syllabifiers from Wang and Narayanan (2007), Räsänen, Doyle, and Frank (2018a), and Räsänen, Seshadri, and Casillas (2018b). Since SylNet systematically showed superior performance to the tested alternatives, we only report ALICE performance, with SylNet as the chosen syllablebased feature extractor.…”

Section: Syllable Count Estimationmentioning

confidence: 99%

See 1 more Smart Citation

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

et al. 2020

Self Cite

View full text Add to dashboard Cite

Recordings captured by wearable microphones are a standard method for investigating young children's language environments. A key measure to quantify from such data is the amount of speech present in children's home environments. To this end, the LENA recorder and software-a popular system for measuring linguistic input-estimates the number of adult words that children may hear over the course of a recording. However, word count estimation is challenging to do in a language-independent manner; the relationship between observable acoustic patterns and language-specific lexical entities is far from uniform across human languages. In this paper, we ask whether some alternative linguistic units, namely phone(me)s or syllables, could be measured instead of, or in parallel with, words in order to achieve improved cross-linguistic applicability and comparability of an automated system for measuring child language input. We discuss the advantages and disadvantages of measuring different units from theoretical and technical points of view. We also investigate the practical applicability of measuring such units using a novel system called Automatic LInguistic unit Count Estimator (ALICE) together with audio from seven child-centered daylong audio corpora from diverse cultural and linguistic environments. We show that language-independent measurement of phoneme counts is somewhat more accurate than syllables or words, but all three are highly correlated with human annotations on the same data. We share an open-source implementation of ALICE for use by the language research community, enabling automatic phoneme, syllable, and word count estimation from child-centered audio recordings.

show abstract

Section: Technical Considerationsmentioning

confidence: 99%

Section: Syllable Count Estimationmentioning

confidence: 99%

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, this downside of "traditional" syllabification algorithms can potentially be tackled with machine learning approaches. By training the system using a large number of syllables from different languages (potentially from different noise conditions), a machine learning algorithm may learn a general solution for syllable counting that better tolerates non-speech audio content than the traditional signal processing approaches (e.g., Räsänen, Seshadri & Casillas, 2018;. This makes syllables a potential candidate unit that could be estimated directly from speech instead of using correlated features such as speech duration or phone counts as linear correlates of syllable counts.…”

Section: Technical Considerationsmentioning

confidence: 99%

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

Räsänen¹,

Seshadri²,

Lavechin³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Recordings captured by wearable microphones are a standard method for investigating young children’s language environments. A key measure to quantify from such data is the amount of speech present in children’s home environments. To this end, the LENA recorder and software—a popular system for measuring linguistic input—estimates the number of adult words that children may hear over the course of a recording. However, word count estimation is challenging to do in a language-independent manner; the relationship between observable acoustic patterns and language-specific lexical entities is far from uniform across human languages. In this paper, we ask whether some alternative linguistic units, namely phone(me)s or syllables, could be measured instead of, or in parallel with, words in order to achieve improved cross-linguistic applicability and comparability of an automated system for measuring child language input. We discuss the advantages and disadvantages of measuring different units from theoretical and technical points of view. We also investigate the practical applicability of measuring such units using a novel system called Automatic LInguistic unit Count Estimator (ALICE) together with audio from seven child-centered daylong audio corpora from diverse cultural and linguistic environments. We show that language-independent measurement of phoneme counts is somewhat more accurate than syllables or words, but all three are highly correlated with human annotations on the same data. We share an open-source implementation of ALICE for use by the language research community, allowing automatic phoneme, syllable, and word count estimation from child-centered audio recordings.

show abstract

“…Okko Rasanen, Shreyas Seshadri et. al [6] presents an Automatic Word Count Estimation (AWCE) to estimate a number of words spoken in an audio recording. But this method is used only for a words that are in an audio format.…”

Section: Related Workmentioning

confidence: 99%

Development of Research Proposal Selection Based on Domain Ontology using K-Means Categorical Clustering

E.*,

M.,

N.J.

2020

IJITEE

View full text Add to dashboard Cite

With the prompt improvement in research progress of various zones, selection of research proposals became a remarkable methodology in many research funding agencies and organizations. When a less number of research proposals are received, then it is ease to cluster the research proposals and the selection process became as non-problematic way. If a number of research proposals elevated, then the clustering and selecting the proposals became complicated. In current system, proposals grouping is done in manual-based or along with their similarities in subject disciplinaries which yield irrelevant results in some cases. The main goal of this research work is to develop an enhanced system in selection of research proposals based on Domain ontology, where the ontology acts as a searching criteria for the topics of research proposals. This proposed system will help to select the topics of research proposals in well-systematic way without the interference of manual progression. In this paper, an algorithm is proposed as Scikit-learn K-means Multiclass Document Clustering(SKMDC) to group each subject discipline according to their sub-topics and sub-domains. Here, the k-means clustering technique is implemented on categorical data to implement the clustering process. As, the categorical data are not able to applied directly in K-means clustering algorithm, the LabelEncoder method is implemented to encode the text data to numerical values and the dimensions of a dataset are reduced using Principal Component Analysis. This paper also overwhelms the weaknesses of k-means technique in specification of cluster number in initial stage. It is done through the determination of optimal number of clusters by using Elbow Curve method and it is cross-validated through Silhouette Score analysis.

show abstract

Comparison of Syllabification Algorithms and Training Strategies for Robust Word Count Estimation across Different Languages and Recording Conditions

Cited by 9 publications

References 16 publications

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings

Development of Research Proposal Selection Based on Domain Ontology using K-Means Categorical Clustering

Contact Info

Product

Resources

About