Search

In this work, we address the classification in topics of utterances in Japanese received by a speech-oriented guidance system operating in a real environment. The implementation of this kind of systems requires the collection and manual labeling of actual user's utterances, which is a costly process. Because of this, we are interested in evaluating the influence of the amount of data for training in the context of topic classification. For this, we compared the performance of a Support Vector Machine and a Maximum Entropy classifier using training data of different sizes. We used actual data collected by the speech-oriented guidance system Takemaru-kun, from adults and children, and also evaluated the effect of automatic speech recognition (ASR) errors in the classification performance. To deal with the shortness of the utterances we proposed to use characters as features, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show an average performance decrease of 4.6% for ASR results of utterances from adults, and 2.8% for children, when reducing the amount of data for training to its 25%; and a classification performance improvement from 92.2% to 94.1% for adults and 87.2% to 88.3% for children, when using character as features instead of words., APSIPA Annual Summit and Conference 2010, December 14-17, 2010, Biopolis, Singapore.

We investigated a speaker verification method that uses non-audible murmur (NAM) segments using newly collected data and obtained several findings that will be useful when speaker verification systems are made in practice. NAM is recorded using a special microphone placed on the surface of the body, so it includes almost no external noise and is hard for other people to hear. By utilizing these properties, we have already reported a text-dependent method using NAM segments that can use a keyword phrase safely. This paper extends the examination with newly collected data consisting of NAM uttered by 18 male and 9 female imposter speakers and by 18 male and 10 female customer speakers. Experiments with various numbers of training utterances and sessions show that it is effective to use data recorded in multiple sessions. We also investigated the minimum number of training utterances needed in our method., INTERSPEECH2007: 8th Annual Conference of the International Speech Communication Association, August 27-31, 2007, Antwerp, Belgium.

We investigate a novel method for speaker verification with nonaudible murmur (NAM) segments. NAM is recorded using a special microphone placed on the neck and is hard for other people to hear. We have already reported a method based on a support vector machine (SVM) using NAM segments to use a keyword phrase effectively. To further exploit keyword-specific features, we introduce a global alignment (GA) kernel and penalized logistic regression machine (PLRM). In the experiments using NAM from 55 speakers, our method achieved an error reduction rate of roughly 60% compared with the SVM-based method using a polynomial kernel., INTERSPEECH2008: 9th Annual Conference of the International Speech Communication Association, September 22-26, 2008, Brisbane, Australia.

This work addresses the classification in topics of utterances in Japanese, received by a speech-oriented guidance system operating in a real environment. For this, we compare the performance of Support Vector Machine and PrefixSpan Boosting, against a conventional Maximum Entropy classification method. We are interested in evaluating their strength against automatic speech recognition (ASR) errors and the sparseness of the features present in spontaneous speech. To deal with the shortness of the utterances, we also proposed to use characters as features instead of words, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show a classification performance improvement from 92.2% to 94.4%, with Support Vector Machine using character unigrams and bigrams as features, in comparison to the conventional method., INTERSPEECH2010: 11th Annual Conference of the International Speech Communication Association, September 26-30, 2010, Chiba, Japan.

Stacked generalization is a method that allows combining output of multiple classifiers using a second-level classification, minimizing the generalization error of first-level classifiers and achieving greater predictive accuracy. In a previous work, we compared the performance of support vector machine (SVM) with radial basis function (RBF) kernel, prefixspan boosting (pboost) and maximum entropy (ME) in the classification in topics of spoken inquiries in Japanese received by a guidance system. In the present work, we employ a stacked generalization scheme that uses predictions of SVM with RBF kernel, pboost and ME as input for a second-level classification using linear SVM. Experimental results show an improvement in performance from 94.2% to 95.1% in the classification of automatic speech recognition (ASR) 1-best results of adults’ inquiries and from 88.3% to 89.2% for children’s inquiries, when using stacked generalization in comparison to the individual performance of the first-level classifiers., APSIPA ASC 2011: 2011 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, October 18-21, 2011, Xi'an, China.

We investigate a discrimination method for invalid and valid spoken inquiries, received by an speech-oriented guidance system operating in a real environment. Invalid spoken inquiries include background voices, which are not directly uttered to the system, and nonsense utterances. Such spoken inquiries should be rejected beforehand. By now, we have reported a method using the likelihood values of Gaussian mixture models (GMMs) to discriminate invalid spoken inquiries from valid ones. In this paper, we improve the performance by utilizing not only the likelihood values but also other information in spoken inquiries such as bag-of-words (BOW), utterance duration, and signal-to-noise ratio (SNR). To deal with these multiple information, we use support vector machine (SVM) with redial basis function (RBF) kernel and maximum entropy (ME) and compare the performance. In the experiments, we archive 86.6% of F-measure for SVM and 84.2% for ME, while F-measure for GMM-based method is 81.7%., INTERSPEECH 2012: The 13th Annual Conference of the International Speech Communication Association, September 9-13, 2012, Portland, Oregon, USA.

We investigate the topic classification of spoken inquiries received by a guidance system with a semi-supervised learning approach based on TSVM. We obtained gains in learning approach based on TSVM. We obtained gains in classification performance for specific topics by using this approach, in comparison to those of its supervised counterpart., IWSDS2012: The 4th International Workshop on Spoken Dialog Systems, November 28-30, 2012, Paris, France

We investigate a discrimination method for invalid and valid inputs based on machine learning using bag-of-words comprised from automatic speech recognition result as a classification feature. Changing the amount of training data, we elucidate that using 3000 of them (approx. 2 weeks of system inputs) shows enough classification performance., IWSDS2012: The 4th International Workshop on Spoken Dialog Systems, November 28-30, 2012, Paris, France

We propose a speaker verification method using non-audible murmur (NAM) segments, which are different from normal speech and hard for other people to catch them. To use NAM, we therefore take a text-dependent verification strategy in which each user utters her/his own keyword phrase and utilize not only speaker-specific but also keyword-specific acoustic information. We expect this strategy to yield a relatively high performance. NAM segments, which consist of multiple short-term feature vectors, are used as input vectors to capture keyword-specific acoustic information well. To handle segments with a large number of dimensions, we use the support vector machine (SVM). In experiments using NAM data of 19 male and 10 female speakers recorded in three different sessions, we achieved equal error rates of 0.04% (male) and 1.1% (female) when using 145-ms-long NAM segments. These rates are half or less those obtained with 25-ms-long input vectors., INTERSPEECH2006: the 9th International Conference on Spoken Language Processing (ICSLP), September 17-21, 2006, Pittsburgh, Pennsylvania, USA.

Example-based question answering (QA) is an effective approach for real-world spoken dialogue systems. A limitation of an example-based QA is that a system cannot appropriately respond to a user’s question, if a similar questionanswer pair does not exist in the question and answer database (QADB). For a robust spoken dialogue system, it is important to classify if a user’s utterance is in the task or out of the task. In this paper, we describe our approach for out-of-task utterance (OOT) detection. Using the Support Vector Machines (SVM), the detection model is trained with the bag of words from the 10-best automatic speech recognition (ASR) results. The number of words in a question, the number of unknown words, and the maximum similarity score against QADB are also used as features for the OOT detection. We apply our detection model to the Takemaru-kun dialogue system. We evaluate our detection model using adult's utterances of two years and child’s utterances of one year spoken to Takemaru-kun. Our proposed method decreases the Equal Error Rate (EER) using speech recognition results by 4.4% (from 21.3% to 16.9%) in adult’s speech and by 3.6% (from 31.8% to 28.2%) in child’s speech, compared with the baseline method., APSIPA ASC 2011: 2011 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, October 18-21, 2011, Xi'an, China.

We propose a new framework that generalizes the parameters of neural network models to $C^*$-algebra-valued ones. $C^*$-algebra is a generalization of the space of complex numbers. A typical example is the space of continuous functions on a compact space. This generalization enables us to combine multiple models continuously and use tools for functions such as regression and integration. Consequently, we can learn features of data efficiently and adapt the models to problems continuously. We apply our framework to practical problems such as density estimation and few-shot learning and show that our framework enables us to learn features of data even with a limited number of samples. Our new framework highlights the potential possibility of applying the theory of $C^*$-algebra to general neural network models., Comment: Accepted at ICML 2022

The COVID-19 pandemic, which began in December 2019, progressed in a complicated manner and thus caused problems worldwide. Seeking clues to the reasons for the complicated progression is necessary but challenging in the fight against the pandemic. We sought clues by investigating the relationship between reactions on social media and the COVID-19 epidemic in Japan. Twitter was selected as the social media platform for study because it has a large user base in Japan and because it quickly propagates short topic-focused messages ("tweets"). Analysis using Japanese Twitter data suggests that reactions on social media and the progression of the COVID-19 pandemic may have a close relationship. Experiments to evaluate the potential of using tweets to support the prediction of how an epidemic will progress demonstrated the value of using epidemic-related social media data. Our findings provide insights into the relationship between user reactions on social media, particularly Twitter, and epidemic progression, which can be used to fight pandemics., Comment: 14 pages, 4 figures

The Japanese version of the Children’s Communication Checklist-2 (CCC-2) was rated by caregivers in a large national population sample of 22,871 children aged 3–15 years. The General Communication Composite (GCC) of the CCC-2 exhibited a distribution with a single-factor structure. The GCC distribution between autism spectrum disorders (ASD) and language impairment (LI) groups in the general population fit inside a bell curve with significant overlap with the general population, and a continuum was evident between groups. No evidence of a natural cutoff that would differentiate categorically affected from unaffected children was seen. The Social Interaction Deviance Composite (SIDC) supported the notion that ASD and LI are on the opposite endpoints of a SIDC continuum of communication impairment. © 2017 Springer Science+Business Media, LLC, Embargo Period 12 months

形態: カラー図版あり, 会議情報: 第5回EFD/CFD融合ワークショップ (2012年1月23日. 秋葉原コンベンションホール), 千代田区, 東京都, Physical characteristics: Original contains color illustrations, Meeting Information: The 5th Workshop on Integration of EFD and CFD (January 23, 2012. AKIHABARA Convention Hall), Chiyoda-ku, Tokyo, Japan

We investigate the topic classification of spoken inquiries received by a guidance system with a semi-supervised learning approach based on TSVM. We obtained gains in learning approach based on TSVM. We obtained gains in classification performance for specific topics by using this approach, in comparison to those of its supervised counterpart.

We investigate a discrimination method for invalid and valid inputs based on machine learning using bag-of-words comprised from automatic speech recognition result as a classification feature. Changing the amount of training data, we elucidate that using 3000 of them (approx. 2 weeks of system inputs) shows enough classification performance.

application/pdf, We investigate a discrimination method for invalid and valid spoken inquiries, received by an speech-oriented guidance system operating in a real environment. Invalid spoken inquiries include background voices, which are not directly uttered to the system, and nonsense utterances. Such spoken inquiries should be rejected beforehand. By now, we have reported a method using the likelihood values of Gaussian mixture models (GMMs) to discriminate invalid spoken inquiries from valid ones. In this paper, we improve the performance by utilizing not only the likelihood values but also other information in spoken inquiries such as bag-of-words (BOW), utterance duration, and signal-to-noise ratio (SNR). To deal with these multiple information, we use support vector machine (SVM) with redial basis function (RBF) kernel and maximum entropy (ME) and compare the performance. In the experiments, we archive 86.6% of F-measure for SVM and 84.2% for ME, while F-measure for GMM-based method is 81.7%.

We investigate a discrimination method for invalid and valid spoken inquiries, received by an speech-oriented guidance system operating in a real environment. Invalid spoken inquiries include background voices, which are not directly uttered to the system, and nonsense utterances. Such spoken inquiries should be rejected beforehand. By now, we have reported a method using the likelihood values of Gaussian mixture models (GMMs) to discriminate invalid spoken inquiries from valid ones. In this paper, we improve the performance by utilizing not only the likelihood values but also other information in spoken inquiries such as bag-of-words (BOW), utterance duration, and signal-to-noise ratio (SNR). To deal with these multiple information, we use support vector machine (SVM) with redial basis function (RBF) kernel and maximum entropy (ME) and compare the performance. In the experiments, we archive 86.6% of F-measure for SVM and 84.2% for ME, while F-measure for GMM-based method is 81.7%., conference paper

We investigate a discrimination method for invalid and valid inputs based on machine learning using bag-of-words comprised from automatic speech recognition result as a classification feature. Changing the amount of training data, we elucidate that using 3000 of them (approx. 2 weeks of system inputs) shows enough classification performance., conference paper

We investigate the topic classification of spoken inquiries received by a guidance system with a semi-supervised learning approach based on TSVM. We obtained gains in learning approach based on TSVM. We obtained gains in classification performance for specific topics by using this approach, in comparison to those of its supervised counterpart., conference paper

Stacked generalization is a method that allows combining output of multiple classifiers using a second-level classification, minimizing the generalization error of first-level classifiers and achieving greater predictive accuracy. In a previous work, we compared the performance of support vector machine (SVM) with radial basis function (RBF) kernel, prefixspan boosting (pboost) and maximum entropy (ME) in the classification in topics of spoken inquiries in Japanese received by a guidance system. In the present work, we employ a stacked generalization scheme that uses predictions of SVM with RBF kernel, pboost and ME as input for a second-level classification using linear SVM. Experimental results show an improvement in performance from 94.2% to 95.1% in the classification of automatic speech recognition (ASR) 1-best results of adults’ inquiries and from 88.3% to 89.2% for children’s inquiries, when using stacked generalization in comparison to the individual performance of the first-level classifiers.

Example-based question answering (QA) is an effective approach for real-world spoken dialogue systems. A limitation of an example-based QA is that a system cannot appropriately respond to a user’s question, if a similar questionanswer pair does not exist in the question and answer database (QADB). For a robust spoken dialogue system, it is important to classify if a user’s utterance is in the task or out of the task. In this paper, we describe our approach for out-of-task utterance (OOT) detection. Using the Support Vector Machines (SVM), the detection model is trained with the bag of words from the 10-best automatic speech recognition (ASR) results. The number of words in a question, the number of unknown words, and the maximum similarity score against QADB are also used as features for the OOT detection. We apply our detection model to the Takemaru-kun dialogue system. We evaluate our detection model using adult's utterances of two years and child’s utterances of one year spoken to Takemaru-kun. Our proposed method decreases the Equal Error Rate (EER) using speech recognition results by 4.4% (from 21.3% to 16.9%) in adult’s speech and by 3.6% (from 31.8% to 28.2%) in child’s speech, compared with the baseline method.

Example-based question answering (QA) is an effective approach for real-world spoken dialogue systems. A limitation of an example-based QA is that a system cannot appropriately respond to a user’s question, if a similar questionanswer pair does not exist in the question and answer database (QADB). For a robust spoken dialogue system, it is important to classify if a user’s utterance is in the task or out of the task. In this paper, we describe our approach for out-of-task utterance (OOT) detection. Using the Support Vector Machines (SVM), the detection model is trained with the bag of words from the 10-best automatic speech recognition (ASR) results. The number of words in a question, the number of unknown words, and the maximum similarity score against QADB are also used as features for the OOT detection. We apply our detection model to the Takemaru-kun dialogue system. We evaluate our detection model using adult's utterances of two years and child’s utterances of one year spoken to Takemaru-kun. Our proposed method decreases the Equal Error Rate (EER) using speech recognition results by 4.4% (from 21.3% to 16.9%) in adult’s speech and by 3.6% (from 31.8% to 28.2%) in child’s speech, compared with the baseline method., conference paper

Stacked generalization is a method that allows combining output of multiple classifiers using a second-level classification, minimizing the generalization error of first-level classifiers and achieving greater predictive accuracy. In a previous work, we compared the performance of support vector machine (SVM) with radial basis function (RBF) kernel, prefixspan boosting (pboost) and maximum entropy (ME) in the classification in topics of spoken inquiries in Japanese received by a guidance system. In the present work, we employ a stacked generalization scheme that uses predictions of SVM with RBF kernel, pboost and ME as input for a second-level classification using linear SVM. Experimental results show an improvement in performance from 94.2% to 95.1% in the classification of automatic speech recognition (ASR) 1-best results of adults’ inquiries and from 88.3% to 89.2% for children’s inquiries, when using stacked generalization in comparison to the individual performance of the first-level classifiers., conference paper

In this work, we address the classification in topics of utterances in Japanese received by a speech-oriented guidance system operating in a real environment. The implementation of this kind of systems requires the collection and manual labeling of actual user's utterances, which is a costly process. Because of this, we are interested in evaluating the influence of the amount of data for training in the context of topic classification. For this, we compared the performance of a Support Vector Machine and a Maximum Entropy classifier using training data of different sizes. We used actual data collected by the speech-oriented guidance system Takemaru-kun, from adults and children, and also evaluated the effect of automatic speech recognition (ASR) errors in the classification performance. To deal with the shortness of the utterances we proposed to use characters as features, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show an average performance decrease of 4.6% for ASR results of utterances from adults, and 2.8% for children, when reducing the amount of data for training to its 25%; and a classification performance improvement from 92.2% to 94.1% for adults and 87.2% to 88.3% for children, when using character as features instead of words.

This work addresses the classification in topics of utterances in Japanese, received by a speech-oriented guidance system operating in a real environment. For this, we compare the performance of Support Vector Machine and PrefixSpan Boosting, against a conventional Maximum Entropy classification method. We are interested in evaluating their strength against automatic speech recognition (ASR) errors and the sparseness of the features present in spontaneous speech. To deal with the shortness of the utterances, we also proposed to use characters as features instead of words, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show a classification performance improvement from 92.2% to 94.4%, with Support Vector Machine using character unigrams and bigrams as features, in comparison to the conventional method.

This work addresses the classification in topics of utterances in Japanese, received by a speech-oriented guidance system operating in a real environment. For this, we compare the performance of Support Vector Machine and PrefixSpan Boosting, against a conventional Maximum Entropy classification method. We are interested in evaluating their strength against automatic speech recognition (ASR) errors and the sparseness of the features present in spontaneous speech. To deal with the shortness of the utterances, we also proposed to use characters as features instead of words, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show a classification performance improvement from 92.2% to 94.4%, with Support Vector Machine using character unigrams and bigrams as features, in comparison to the conventional method., conference paper

In this work, we address the classification in topics of utterances in Japanese received by a speech-oriented guidance system operating in a real environment. The implementation of this kind of systems requires the collection and manual labeling of actual user's utterances, which is a costly process. Because of this, we are interested in evaluating the influence of the amount of data for training in the context of topic classification. For this, we compared the performance of a Support Vector Machine and a Maximum Entropy classifier using training data of different sizes. We used actual data collected by the speech-oriented guidance system Takemaru-kun, from adults and children, and also evaluated the effect of automatic speech recognition (ASR) errors in the classification performance. To deal with the shortness of the utterances we proposed to use characters as features, which is possible with the Japanese language due to the presence of kanji; ideograms from Chinese characters that represent not only sound but meaning. Experimental results show an average performance decrease of 4.6% for ASR results of utterances from adults, and 2.8% for children, when reducing the amount of data for training to its 25%; and a classification performance improvement from 92.2% to 94.1% for adults and 87.2% to 88.3% for children, when using character as features instead of words., conference paper

We investigate a novel method for speaker verification with nonaudible murmur (NAM) segments. NAM is recorded using a special microphone placed on the neck and is hard for other people to hear. We have already reported a method based on a support vector machine (SVM) using NAM segments to use a keyword phrase effectively. To further exploit keyword-specific features, we introduce a global alignment (GA) kernel and penalized logistic regression machine (PLRM). In the experiments using NAM from 55 speakers, our method achieved an error reduction rate of roughly 60% compared with the SVM-based method using a polynomial kernel.

We investigate a novel method for speaker verification with nonaudible murmur (NAM) segments. NAM is recorded using a special microphone placed on the neck and is hard for other people to hear. We have already reported a method based on a support vector machine (SVM) using NAM segments to use a keyword phrase effectively. To further exploit keyword-specific features, we introduce a global alignment (GA) kernel and penalized logistic regression machine (PLRM). In the experiments using NAM from 55 speakers, our method achieved an error reduction rate of roughly 60% compared with the SVM-based method using a polynomial kernel., conference paper

One significant problem for spoken language systems is how to cope with users' out-of-domain (OOD) utterances which cannot be handled by the back-end application system. In this paper, we propose a novel OOD detection framework, which makes use of the classification confidence scores of multiple topics and applies a linear discriminant model to perform in-domain verification. The verification model is trained using a combination of deleted interpolation of the in-domain data and minimum-classification-error training, and does not require actual OOD data during the training process, thus realizing high portability. When applied to the "phrasebook" system, a single utterance read-style speech task, the proposed approach achieves an absolute reduction in OOD detection errors of up to 8.1 points (40% relative) compared to a baseline method based on the maximum topic classification score. Furthermore, the proposed approach realizes comparable performance to an equivalent system trained on both in-domain and OOD data, while requiring no OOD data during training. We also apply this framework to the "machine-aided-dialogue" corpus, a spontaneous dialogue speech task, and extend the framework in two manners. First, we introduce topic clustering which enables reliable topic confidence scores to be generated even for indistinct utterances, and second, we implement methods to effectively incorporate dialogue context. Integration of these two methods into the proposed framework significantly improves OOD detection performance, achieving a further reduction in equal error rate (EER) of 7.9 points.

We investigated a speaker verification method that uses non-audible murmur (NAM) segments using newly collected data and obtained several findings that will be useful when speaker verification systems are made in practice. NAM is recorded using a special microphone placed on the surface of the body, so it includes almost no external noise and is hard for other people to hear. By utilizing these properties, we have already reported a text-dependent method using NAM segments that can use a keyword phrase safely. This paper extends the examination with newly collected data consisting of NAM uttered by 18 male and 9 female imposter speakers and by 18 male and 10 female customer speakers. Experiments with various numbers of training utterances and sessions show that it is effective to use data recorded in multiple sessions. We also investigated the minimum number of training utterances needed in our method.

We investigated a speaker verification method that uses non-audible murmur (NAM) segments using newly collected data and obtained several findings that will be useful when speaker verification systems are made in practice. NAM is recorded using a special microphone placed on the surface of the body, so it includes almost no external noise and is hard for other people to hear. By utilizing these properties, we have already reported a text-dependent method using NAM segments that can use a keyword phrase safely. This paper extends the examination with newly collected data consisting of NAM uttered by 18 male and 9 female imposter speakers and by 18 male and 10 female customer speakers. Experiments with various numbers of training utterances and sessions show that it is effective to use data recorded in multiple sessions. We also investigated the minimum number of training utterances needed in our method., conference paper

We propose a speaker verification method using non-audible murmur (NAM) segments, which are different from normal speech and hard for other people to catch them. To use NAM, we therefore take a text-dependent verification strategy in which each user utters her/his own keyword phrase and utilize not only speaker-specific but also keyword-specific acoustic information. We expect this strategy to yield a relatively high performance. NAM segments, which consist of multiple short-term feature vectors, are used as input vectors to capture keyword-specific acoustic information well. To handle segments with a large number of dimensions, we use the support vector machine (SVM). In experiments using NAM data of 19 male and 10 female speakers recorded in three different sessions, we achieved equal error rates of 0.04% (male) and 1.1% (female) when using 145-ms-long NAM segments. These rates are half or less those obtained with 25-ms-long input vectors.

We propose a speaker verification method using non-audible murmur (NAM) segments, which are different from normal speech and hard for other people to catch them. To use NAM, we therefore take a text-dependent verification strategy in which each user utters her/his own keyword phrase and utilize not only speaker-specific but also keyword-specific acoustic information. We expect this strategy to yield a relatively high performance. NAM segments, which consist of multiple short-term feature vectors, are used as input vectors to capture keyword-specific acoustic information well. To handle segments with a large number of dimensions, we use the support vector machine (SVM). In experiments using NAM data of 19 male and 10 female speakers recorded in three different sessions, we achieved equal error rates of 0.04% (male) and 1.1% (female) when using 145-ms-long NAM segments. These rates are half or less those obtained with 25-ms-long input vectors., conference paper

We propose in this paper a new family of kernels to handle times series, notably speech data, within the framework of kernel methods which includes popular algorithms such as the Support Vector Machine. These kernels elaborate on the well known Dynamic Time Warping (DTW) family of distances by considering the same set of elementary operations, namely substitutions and repetitions of tokens, to map a sequence onto another. Associating to each of these operations a given score, DTW algorithms use dynamic programming techniques to compute an optimal sequence of operations with high overall score. In this paper we consider instead the score spanned by all possible alignments, take a smoothed version of their maximum and derive a kernel out of this formulation. We prove that this kernel is positive definite under favorable conditions and show how it can be tuned effectively for practical applications as we report encouraging results on a speech recognition task.

The performance of automatic speech recognition system is not enough to perfectly recognize speech and users must rephrase their input utterances when recognition errors occur. However, speaking styles of error recovery utterances are significantly changed from that of normal utterances. Because of these changes, the performance of speech recognition system is degraded to the contrary....

The performance of automatic speech recognition system is not enough to perfectly recognize speech and users must rephrase their input utterances when recognition errors occur. However, speaking styles of error recovery utterances are significantly changed from that of normal utterances. Because of these changes, the performance of speech recognition system is degraded to the contrary....