Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846319
Leimin Tian, Johanna D. Moore, Catherine Lai
Automatic emotion recognition is vital for building natural and engaging human-computer interaction systems. Combining information from multiple modalities typically improves emotion recognition performance. In previous work, features from different modalities have generally been fused at the same level with two types of fusion strategies: Feature-Level fusion, which concatenates feature sets before recognition; and Decision-Level fusion, which makes the final decision based on outputs of the unimodal models. However, different features may describe data at different time scales or have different levels of abstraction. Cognitive Science research also indicates that when perceiving emotions, humans use information from different modalities at different cognitive levels and time steps. Therefore, we propose a Hierarchical fusion strategy for multimodal emotion recognition, which incorporates global or more abstract features at higher levels of its knowledge-inspired structure. We build multimodal emotion recognition models combining state-of-the-art acoustic and lexical features to study the performance of the proposed Hierarchical fusion. Experiments on two emotion databases of spoken dialogue show that this fusion strategy consistently outperforms both Feature-Level and Decision-Level fusion. The multimodal emotion recognition models using the Hierarchical fusion strategy achieved state-of-the-art performance on recognizing emotions in both spontaneous and acted dialogue.
{"title":"Recognizing emotions in spoken dialogue with hierarchically fused acoustic and lexical features","authors":"Leimin Tian, Johanna D. Moore, Catherine Lai","doi":"10.1109/SLT.2016.7846319","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846319","url":null,"abstract":"Automatic emotion recognition is vital for building natural and engaging human-computer interaction systems. Combining information from multiple modalities typically improves emotion recognition performance. In previous work, features from different modalities have generally been fused at the same level with two types of fusion strategies: Feature-Level fusion, which concatenates feature sets before recognition; and Decision-Level fusion, which makes the final decision based on outputs of the unimodal models. However, different features may describe data at different time scales or have different levels of abstraction. Cognitive Science research also indicates that when perceiving emotions, humans use information from different modalities at different cognitive levels and time steps. Therefore, we propose a Hierarchical fusion strategy for multimodal emotion recognition, which incorporates global or more abstract features at higher levels of its knowledge-inspired structure. We build multimodal emotion recognition models combining state-of-the-art acoustic and lexical features to study the performance of the proposed Hierarchical fusion. Experiments on two emotion databases of spoken dialogue show that this fusion strategy consistently outperforms both Feature-Level and Decision-Level fusion. The multimodal emotion recognition models using the Hierarchical fusion strategy achieved state-of-the-art performance on recognizing emotions in both spontaneous and acted dialogue.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124895494","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846334
Tomohiro Tanaka, Takafumi Moriya, T. Shinozaki, Shinji Watanabe, Takaaki Hori, Kevin Duh
Long short-term memory (LSTM) recurrent neural network based language models are known to improve speech recognition performance. However, significant effort is required to optimize network structures and training configurations. In this study, we automate the development process using evolutionary algorithms. In particular, we apply the covariance matrix adaptation-evolution strategy (CMA-ES), which has demonstrated robustness in other black box hyper-parameter optimization problems. By flexibly allowing optimization of various meta-parameters including layer wise unit types, our method automatically finds a configuration that gives improved recognition performance. Further, by using a Pareto based multi-objective CMA-ES, both WER and computational time were reduced jointly: after 10 generations, relative WER and computational time reductions for decoding were 4.1% and 22.7% respectively, compared to an initial baseline system whose WER was 8.7%.
{"title":"Automated structure discovery and parameter tuning of neural network language model based on evolution strategy","authors":"Tomohiro Tanaka, Takafumi Moriya, T. Shinozaki, Shinji Watanabe, Takaaki Hori, Kevin Duh","doi":"10.1109/SLT.2016.7846334","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846334","url":null,"abstract":"Long short-term memory (LSTM) recurrent neural network based language models are known to improve speech recognition performance. However, significant effort is required to optimize network structures and training configurations. In this study, we automate the development process using evolutionary algorithms. In particular, we apply the covariance matrix adaptation-evolution strategy (CMA-ES), which has demonstrated robustness in other black box hyper-parameter optimization problems. By flexibly allowing optimization of various meta-parameters including layer wise unit types, our method automatically finds a configuration that gives improved recognition performance. Further, by using a Pareto based multi-objective CMA-ES, both WER and computational time were reduced jointly: after 10 generations, relative WER and computational time reductions for decoding were 4.1% and 22.7% respectively, compared to an initial baseline system whose WER was 8.7%.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114425383","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846302
A. Michaely, M. Ghodsi, Zelin Wu, Justin Scheiner, Petar S. Aleksic
It has been shown in the literature that automatic speech recognition systems can greatly benefit from contextual information [1, 2, 3, 4, 5]. Contextual information can be used to simplify the beam search and improve recognition accuracy. Types of useful contextual information can include the name of the application the user is in, the contents of the user's phone screen, the user's location, a certain dialog state, etc. Building a separate language model for each of these types of context is not feasible due to limited resources or limited amounts of training data. In this paper we describe an approach for unsupervised learning of contextual information and automatic building of contextual biasing models. Our approach can be used to build a large number of small contextual models from a limited amount of available unsupervised training data. We describe how n-grams relevant for a particular context are automatically selected as well as how an optimal size of a final contextual model is chosen. Our experimental results show great accuracy improvements for several types of context.
{"title":"Unsupervised context learning for speech recognition","authors":"A. Michaely, M. Ghodsi, Zelin Wu, Justin Scheiner, Petar S. Aleksic","doi":"10.1109/SLT.2016.7846302","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846302","url":null,"abstract":"It has been shown in the literature that automatic speech recognition systems can greatly benefit from contextual information [1, 2, 3, 4, 5]. Contextual information can be used to simplify the beam search and improve recognition accuracy. Types of useful contextual information can include the name of the application the user is in, the contents of the user's phone screen, the user's location, a certain dialog state, etc. Building a separate language model for each of these types of context is not feasible due to limited resources or limited amounts of training data. In this paper we describe an approach for unsupervised learning of contextual information and automatic building of contextual biasing models. Our approach can be used to build a large number of small contextual models from a limited amount of available unsupervised training data. We describe how n-grams relevant for a particular context are automatically selected as well as how an optimal size of a final contextual model is chosen. Our experimental results show great accuracy improvements for several types of context.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116112553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846298
Yike Zhang, Pengyuan Zhang, Ta Li, Yonghong Yan
The vocabulary is a vital component of automatic speech recognition(ASR) systems. For a specific Chinese speech recognition task, using a large general vocabulary not only leads to a much longer time to decode, but also hurts the recognition accuracy. In this paper, we proposed an unsupervised algorithm to select task-specific words from a large general vocabulary. The out-of-vocabulary(OOV) rate is a measure of vocabularies, and it is related to the recognition accuracy. However, it is hard to compute OOV rate for a Chinese vocabulary, since OOVs are often segmented into single Chinese characters and most Chinese vocabularies contain all the single Chinese characters. To deal with this problem, we proposed a novel method to estimate the OOV rate of Chinese vocabularies. In experiments, we found that our estimated OOV rate is related to the character error rate(CER) of recognition. Our proposed vocabulary selection method provided both the lowest OOV rate and CER on two Chinese conversational telephone speech(CTS) evaluation sets compared to the general vocabulary and frequency based vocabulary selection method. In addition, our proposed method significantly reduced the size of the language model(LM) and the corresponding weighted finite state transducer(WFST) network, which led to a more efficient decoding.
{"title":"An unsupervised vocabulary selection technique for Chinese automatic speech recognition","authors":"Yike Zhang, Pengyuan Zhang, Ta Li, Yonghong Yan","doi":"10.1109/SLT.2016.7846298","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846298","url":null,"abstract":"The vocabulary is a vital component of automatic speech recognition(ASR) systems. For a specific Chinese speech recognition task, using a large general vocabulary not only leads to a much longer time to decode, but also hurts the recognition accuracy. In this paper, we proposed an unsupervised algorithm to select task-specific words from a large general vocabulary. The out-of-vocabulary(OOV) rate is a measure of vocabularies, and it is related to the recognition accuracy. However, it is hard to compute OOV rate for a Chinese vocabulary, since OOVs are often segmented into single Chinese characters and most Chinese vocabularies contain all the single Chinese characters. To deal with this problem, we proposed a novel method to estimate the OOV rate of Chinese vocabularies. In experiments, we found that our estimated OOV rate is related to the character error rate(CER) of recognition. Our proposed vocabulary selection method provided both the lowest OOV rate and CER on two Chinese conversational telephone speech(CTS) evaluation sets compared to the general vocabulary and frequency based vocabulary selection method. In addition, our proposed method significantly reduced the size of the language model(LM) and the corresponding weighted finite state transducer(WFST) network, which led to a more efficient decoding.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"99 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123135691","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846295
L. Velikovich
This paper introduces a semantic tagger that inserts tags into a word lattice, such as one produced by a real-time large-vocabulary speech recognition system. Benefits of such a tagger include the ability to rescore speech recognition hypotheses based on this metadata, as well as providing rich annotations to clients downstream. We focus on the domain of spoken search queries and voice commands, which can be useful for building an intelligent assistant. We explore a method to distill a pre-existing very large named entity disambiguation (NED) model into a lightweight tagger. This is accomplished by constructing a joint distribution of tagged n-grams from a supervised training corpus, then deriving a conditional distribution for a given lattice. With 300 tagging categories, the tagger achieves a precision of 88.2% and recall of 93.1% on 1-best paths in speech recognition lattices with 2.8ms median latency.
{"title":"Semantic model for fast tagging of word lattices","authors":"L. Velikovich","doi":"10.1109/SLT.2016.7846295","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846295","url":null,"abstract":"This paper introduces a semantic tagger that inserts tags into a word lattice, such as one produced by a real-time large-vocabulary speech recognition system. Benefits of such a tagger include the ability to rescore speech recognition hypotheses based on this metadata, as well as providing rich annotations to clients downstream. We focus on the domain of spoken search queries and voice commands, which can be useful for building an intelligent assistant. We explore a method to distill a pre-existing very large named entity disambiguation (NED) model into a lightweight tagger. This is accomplished by constructing a joint distribution of tagged n-grams from a supervised training corpus, then deriving a conditional distribution for a given lattice. With 300 tagging categories, the tagger achieves a precision of 88.2% and recall of 93.1% on 1-best paths in speech recognition lattices with 2.8ms median latency.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"197 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130033847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846282
Lea Schonherr, Dennis Orth, M. Heckmann, D. Kolossa
To improve the accuracy of audio-visual speaker identification, we propose a new approach, which achieves an optimal combination of the different modalities on the score level. We use the i-vector method for the acoustics and the local binary pattern (LBP) for the visual speaker recognition. Regarding the input data of both modalities, multiple confidence measures are utilized to calculate an optimal weight for the fusion. Thus, oracle weights are chosen in such a way as to maximize the difference between the score of the genuine speaker and the person with the best competing score. Based on these oracle weights a mapping function for weight estimation is learned. To test the approach, various combinations of noise levels for the acoustic and visual data are considered. We show that the weighted multimodal identification is far less influenced by the presence of noise or distortions in acoustic or visual observations in comparison to an unweighted combination.
{"title":"Environmentally robust audio-visual speaker identification","authors":"Lea Schonherr, Dennis Orth, M. Heckmann, D. Kolossa","doi":"10.1109/SLT.2016.7846282","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846282","url":null,"abstract":"To improve the accuracy of audio-visual speaker identification, we propose a new approach, which achieves an optimal combination of the different modalities on the score level. We use the i-vector method for the acoustics and the local binary pattern (LBP) for the visual speaker recognition. Regarding the input data of both modalities, multiple confidence measures are utilized to calculate an optimal weight for the fusion. Thus, oracle weights are chosen in such a way as to maximize the difference between the score of the genuine speaker and the person with the best competing score. Based on these oracle weights a mapping function for weight estimation is learned. To test the approach, various combinations of noise levels for the acoustic and visual data are considered. We show that the weighted multimodal identification is far less influenced by the presence of noise or distortions in acoustic or visual observations in comparison to an unweighted combination.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"142 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124626342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846308
Ivan Kukanov, Ville Hautamäki, S. Siniscalchi, Kehuang Li
In this work, we are interested in boosting speech attribute detection by formulating it as a multi-label classification task, and deep neural networks (DNNs) are used to design speech attribute detectors. A straightforward way to tackle the speech attribute detection task is to estimate DNN parameters using the mean squared error (MSE) loss function and employ a sigmoid function in the DNN output nodes. A more principled way is nonetheless to incorporate the micro-F1 measure, which is a widely used metric in the multi-label classification, into the DNN loss function to directly improve the metric of interest at training time. Micro-F1 is not differentiable, yet we overcome such a problem by casting our task under the maximal figure-of-merit (MFoM) learning framework. The results demonstrate that our MFoM approach consistently outperforms the baseline systems.
{"title":"Deep learning with maximal figure-of-merit cost to advance multi-label speech attribute detection","authors":"Ivan Kukanov, Ville Hautamäki, S. Siniscalchi, Kehuang Li","doi":"10.1109/SLT.2016.7846308","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846308","url":null,"abstract":"In this work, we are interested in boosting speech attribute detection by formulating it as a multi-label classification task, and deep neural networks (DNNs) are used to design speech attribute detectors. A straightforward way to tackle the speech attribute detection task is to estimate DNN parameters using the mean squared error (MSE) loss function and employ a sigmoid function in the DNN output nodes. A more principled way is nonetheless to incorporate the micro-F1 measure, which is a widely used metric in the multi-label classification, into the DNN loss function to directly improve the metric of interest at training time. Micro-F1 is not differentiable, yet we overcome such a problem by casting our task under the maximal figure-of-merit (MFoM) learning framework. The results demonstrate that our MFoM approach consistently outperforms the baseline systems.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121625188","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846268
Mohamed Bouaziz, Mohamed Morchid, Richard Dufour, G. Linarès, R. Mori
Recently, machine learning methods have provided a broad spectrum of original and efficient algorithms based on Deep Neural Networks (DNN) to automatically predict an outcome with respect to a sequence of inputs. Recurrent hidden cells allow these DNN-based models to manage long-term dependencies such as Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM). Nevertheless, these RNNs process a single input stream in one (LSTM) or two (Bidirectional LSTM) directions. But most of the information available nowadays is from multistreams or multimedia documents, and require RNNs to process these information synchronously during the training. This paper presents an original LSTM-based architecture, named Parallel LSTM (PLSTM), that carries out multiple parallel synchronized input sequences in order to predict a common output. The proposed PLSTM method could be used for parallel sequence classification purposes. The PLSTM approach is evaluated on an automatic telecast genre sequences classification task and compared with different state-of-the-art architectures. Results show that the proposed PLSTM method outperforms the baseline n-gram models as well as the state-of-the-art LSTM approach.
{"title":"Parallel Long Short-Term Memory for multi-stream classification","authors":"Mohamed Bouaziz, Mohamed Morchid, Richard Dufour, G. Linarès, R. Mori","doi":"10.1109/SLT.2016.7846268","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846268","url":null,"abstract":"Recently, machine learning methods have provided a broad spectrum of original and efficient algorithms based on Deep Neural Networks (DNN) to automatically predict an outcome with respect to a sequence of inputs. Recurrent hidden cells allow these DNN-based models to manage long-term dependencies such as Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM). Nevertheless, these RNNs process a single input stream in one (LSTM) or two (Bidirectional LSTM) directions. But most of the information available nowadays is from multistreams or multimedia documents, and require RNNs to process these information synchronously during the training. This paper presents an original LSTM-based architecture, named Parallel LSTM (PLSTM), that carries out multiple parallel synchronized input sequences in order to predict a common output. The proposed PLSTM method could be used for parallel sequence classification purposes. The PLSTM approach is evaluated on an automatic telecast genre sequences classification task and compared with different state-of-the-art architectures. Results show that the proposed PLSTM method outperforms the baseline n-gram models as well as the state-of-the-art LSTM approach.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126700902","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846238
Scott Novotney, D. Karakos, J. Silovský, R. Schwartz
We describe our submission to the NIST OpenSAD evaluation of speech activity detection of noisy audio generated by the DARPA RATS program. With frequent transmission degradation, channel interference and other noises added, simple energy thresholds do a poor job at SAD for this audio. The evaluation measured performance on both in-training and novel channels. Our approach used a system combination of feed-forward neural networks and bidirectional LSTM recurrent neural networks. System combination and unsupervised adaptation provided further gains on novel channels that lack training data. These improvements lead to a 26% relative improvement for novel channels over simple decoding. Our system resulted in the lowest error rate on the in-training channels and second on the out-of-training channels.
{"title":"BBN technologies' OpenSAD system","authors":"Scott Novotney, D. Karakos, J. Silovský, R. Schwartz","doi":"10.1109/SLT.2016.7846238","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846238","url":null,"abstract":"We describe our submission to the NIST OpenSAD evaluation of speech activity detection of noisy audio generated by the DARPA RATS program. With frequent transmission degradation, channel interference and other noises added, simple energy thresholds do a poor job at SAD for this audio. The evaluation measured performance on both in-training and novel channels. Our approach used a system combination of feed-forward neural networks and bidirectional LSTM recurrent neural networks. System combination and unsupervised adaptation provided further gains on novel channels that lack training data. These improvements lead to a 26% relative improvement for novel channels over simple decoding. Our system resulted in the lowest error rate on the in-training channels and second on the out-of-training channels.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"21 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125855402","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-12-01DOI: 10.1109/SLT.2016.7846303
Akshay Chandrashekaran, Ian Lane
In this paper, we explore the usage of automated hyper-parameter optimization techniques with scalarization of multiple objectives to find decoder hyper-parameters suitable for a given acoustic and language model for an LVCSR task. We compare manual optimization, random sampling, tree of Parzen estimators, Bayesian Optimization, and genetic algorithm to find a technique that yields better performance than manual optimization in a comparable number of hyper-parameter evaluations. We focus on a scalar combination of word error rate (WER), log of real time factor (logRTF), and peak memory usage, formulated using the augmented Tchebyscheff function(ATF), as the objective function for the automated techniques. For this task, with a constraint on the maximum number of objective evaluations, we find that the best automated optimization technique: Bayesian Optimization outperforms manual optimization by 8% in terms of ATF. We find that memory usage was not a very useful distinguishing factor between different hyper-parameter settings, with trade-offs occurring between RTF and WER a majority of the time. We also try to perform optimization of WER with a hard constraint on the real time factor of 0.1. In this case, performing constrained Bayesian Optimization yields a model that provides an improvement of 2.7% over the best model obtained from manual optimization with 60% the number of evaluations.
{"title":"Automated optimization of decoder hyper-parameters for online LVCSR","authors":"Akshay Chandrashekaran, Ian Lane","doi":"10.1109/SLT.2016.7846303","DOIUrl":"https://doi.org/10.1109/SLT.2016.7846303","url":null,"abstract":"In this paper, we explore the usage of automated hyper-parameter optimization techniques with scalarization of multiple objectives to find decoder hyper-parameters suitable for a given acoustic and language model for an LVCSR task. We compare manual optimization, random sampling, tree of Parzen estimators, Bayesian Optimization, and genetic algorithm to find a technique that yields better performance than manual optimization in a comparable number of hyper-parameter evaluations. We focus on a scalar combination of word error rate (WER), log of real time factor (logRTF), and peak memory usage, formulated using the augmented Tchebyscheff function(ATF), as the objective function for the automated techniques. For this task, with a constraint on the maximum number of objective evaluations, we find that the best automated optimization technique: Bayesian Optimization outperforms manual optimization by 8% in terms of ATF. We find that memory usage was not a very useful distinguishing factor between different hyper-parameter settings, with trade-offs occurring between RTF and WER a majority of the time. We also try to perform optimization of WER with a hard constraint on the real time factor of 0.1. In this case, performing constrained Bayesian Optimization yields a model that provides an improvement of 2.7% over the best model obtained from manual optimization with 60% the number of evaluations.","PeriodicalId":281635,"journal":{"name":"2016 IEEE Spoken Language Technology Workshop (SLT)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125964616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：