Accessibility Automatic Speech Recognition Computer-assisted transcription Confidence measures Docencia en Red Education language model adaptation Language Modeling Language Technologies Length modelling Log-linear models Machine Translation Massive Adaptation Modelat de la longitud Models basats en seqüències de paraules Models log-lineals Multilingualism Neural Machine Translation Opencast Matterhorn Polimedia Sliding window Speaker adaptation Speech Recognition Speech Translation Statistical machine translation streaming text-to-speech transcripciones video lecture repositories Video Lectures
2021 |
Jorge, Javier ; Giménez, Adrià ; Silvestre-Cerdà, Joan Albert ; Civera, Jorge ; Sanchis, Albert ; Alfons, Juan Live Streaming Speech Recognition Using Deep Bidirectional LSTM Acoustic Models and Interpolated Language Models Journal Article IEEE/ACM Transactions on Audio, Speech, and Language Processing, 30 , pp. 148–161, 2021. Abstract | Links | BibTeX | Tags: acoustic modelling, Automatic Speech Recognition, decoding, language modelling, neural networks, streaming @article{Jorge2021b, title = {Live Streaming Speech Recognition Using Deep Bidirectional LSTM Acoustic Models and Interpolated Language Models}, author = {Jorge, Javier and Giménez, Adrià and Silvestre-Cerdà, Joan Albert and Civera, Jorge and Sanchis, Albert and Juan Alfons}, doi = {10.1109/TASLP.2021.3133216}, year = {2021}, date = {2021-11-23}, journal = {IEEE/ACM Transactions on Audio, Speech, and Language Processing}, volume = {30}, pages = {148--161}, abstract = {Although Long-Short Term Memory (LSTM) networks and deep Transformers are now extensively used in offline ASR, it is unclear how best offline systems can be adapted to work with them under the streaming setup. After gaining considerable experience in this regard in recent years, in this paper we show how an optimized, low-latency streaming decoder can be built in which bidirectional LSTM acoustic models, together with general interpolated language models, can be nicely integrated with minimal perfomance degradation. In brief, our streaming decoder consists of a one-pass, real-time search engine relying on a limited-duration window sliding over time and a number of ad hoc acoustic and language model pruning techniques. Extensive empirical assessment is provided on truly streaming tasks derived from the well-known LibriSpeech and TED talks datasets, as well as from TV shows from a large Spanish broadcasting station.}, keywords = {acoustic modelling, Automatic Speech Recognition, decoding, language modelling, neural networks, streaming}, pubstate = {published}, tppubtype = {article} } Although Long-Short Term Memory (LSTM) networks and deep Transformers are now extensively used in offline ASR, it is unclear how best offline systems can be adapted to work with them under the streaming setup. After gaining considerable experience in this regard in recent years, in this paper we show how an optimized, low-latency streaming decoder can be built in which bidirectional LSTM acoustic models, together with general interpolated language models, can be nicely integrated with minimal perfomance degradation. In brief, our streaming decoder consists of a one-pass, real-time search engine relying on a limited-duration window sliding over time and a number of ad hoc acoustic and language model pruning techniques. Extensive empirical assessment is provided on truly streaming tasks derived from the well-known LibriSpeech and TED talks datasets, as well as from TV shows from a large Spanish broadcasting station. |
Pérez, Alejandro; Garcés Díaz-Munío, Gonçal ; Giménez, Adrià; Silvestre-Cerdà, Joan Albert ; Sanchis, Albert; Civera, Jorge; Jiménez, Manuel; Turró, Carlos; Juan, Alfons Towards cross-lingual voice cloning in higher education Journal Article Engineering Applications of Artificial Intelligence, 105 , pp. 104413, 2021. Abstract | Links | BibTeX | Tags: cross-lingual voice conversion, educational resources, multilinguality, OER, text-to-speech @article{Pérez2021, title = {Towards cross-lingual voice cloning in higher education}, author = {Alejandro Pérez and Garcés Díaz-Munío, Gonçal and Adrià Giménez and Silvestre-Cerdà, Joan Albert and Albert Sanchis and Jorge Civera and Manuel Jiménez and Carlos Turró and Alfons Juan}, url = {https://doi.org/10.1016/j.engappai.2021.104413}, year = {2021}, date = {2021-10-01}, journal = {Engineering Applications of Artificial Intelligence}, volume = {105}, pages = {104413}, abstract = {The rapid progress of modern AI tools for automatic speech recognition and machine translation is leading to a progressive cost reduction to produce publishable subtitles for educational videos in multiple languages. Similarly, text-to-speech technology is experiencing large improvements in terms of quality, flexibility and capabilities. In particular, state-of-the-art systems are now capable of seamlessly dealing with multiple languages and speakers in an integrated manner, thus enabling lecturer's voice cloning in languages she/he might not even speak. This work is to report the experience gained on using such systems at the Universitat Politècnica de València (UPV), mainly as a guidance for other educational organizations willing to conduct similar studies. It builds on previous work on the UPV's main repository of educational videos, MediaUPV, to produce multilingual subtitles at scale and low cost. Here, a detailed account is given on how this work has been extended to also allow for massive machine dubbing of MediaUPV. This includes collecting 59 hours of clean speech data from UPV’s academic staff, and extending our production pipeline of subtitles with a state-of-the-art multilingual and multi-speaker text-to-speech system trained from the collected data. Our main result comes from an extensive, subjective evaluation of this system by lecturers contributing to data collection. In brief, it is shown that text-to-speech technology is not only mature enough for its application to MediaUPV, but also needed as soon as possible by students to improve its accessibility and bridge language barriers.}, keywords = {cross-lingual voice conversion, educational resources, multilinguality, OER, text-to-speech}, pubstate = {published}, tppubtype = {article} } The rapid progress of modern AI tools for automatic speech recognition and machine translation is leading to a progressive cost reduction to produce publishable subtitles for educational videos in multiple languages. Similarly, text-to-speech technology is experiencing large improvements in terms of quality, flexibility and capabilities. In particular, state-of-the-art systems are now capable of seamlessly dealing with multiple languages and speakers in an integrated manner, thus enabling lecturer's voice cloning in languages she/he might not even speak. This work is to report the experience gained on using such systems at the Universitat Politècnica de València (UPV), mainly as a guidance for other educational organizations willing to conduct similar studies. It builds on previous work on the UPV's main repository of educational videos, MediaUPV, to produce multilingual subtitles at scale and low cost. Here, a detailed account is given on how this work has been extended to also allow for massive machine dubbing of MediaUPV. This includes collecting 59 hours of clean speech data from UPV’s academic staff, and extending our production pipeline of subtitles with a state-of-the-art multilingual and multi-speaker text-to-speech system trained from the collected data. Our main result comes from an extensive, subjective evaluation of this system by lecturers contributing to data collection. In brief, it is shown that text-to-speech technology is not only mature enough for its application to MediaUPV, but also needed as soon as possible by students to improve its accessibility and bridge language barriers. |
Jorge, Javier; Giménez, Adrià; Baquero-Arnal, Pau; Iranzo-Sánchez, Javier; Pérez-González-de-Martos, Alejandro; Garcés Díaz-Munío, Gonçal V; Silvestre-Cerdà, Joan Albert; Civera, Jorge; Sanchis, Albert; Juan, Alfons MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge Inproceedings Proc. of IberSPEECH 2021, pp. 118–122, Valladolid (Spain), 2021. Abstract | Links | BibTeX | Tags: Automatic Speech Recognition, Natural Language Processing, streaming @inproceedings{Jorge2021, title = {MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge}, author = {Javier Jorge and Adrià Giménez and Pau Baquero-Arnal and Javier Iranzo-Sánchez and Alejandro Pérez-González-de-Martos and Garcés Díaz-Munío, Gonçal V. and Joan Albert Silvestre-Cerdà and Jorge Civera and Albert Sanchis and Alfons Juan}, doi = {10.21437/IberSPEECH.2021-25}, year = {2021}, date = {2021-03-24}, booktitle = {Proc. of IberSPEECH 2021}, pages = {118--122}, address = {Valladolid (Spain)}, abstract = {1st place in IberSpeech-RTVE 2020 TV Speech-to-Text Challenge. [EN] This paper describes the automatic speech recognition (ASR) systems built by the MLLP-VRAIN research group of Universitat Politecnica de València for the Albayzin-RTVE 2020 Speech-to-Text Challenge. The primary system (p-streaming_1500ms_nlt) was a hybrid BLSTM-HMM ASR system using streaming one-pass decoding with a context window of 1.5 seconds and a linear combination of an n-gram, a LSTM, and a Transformer language model (LM). The acoustic model was trained on nearly 4,000 hours of speech data from different sources, using the MLLP's transLectures-UPV toolkit (TLK) and TensorFlow; whilst LMs were trained using SRILM (n-gram), CUED-RNNLM (LSTM) and Fairseq (Transformer), with up to 102G tokens. This system achieved 11.6% and 16.0% WER on the test-2018 and test-2020 sets, respectively. As it is streaming-enabled, it could be put into production environments for automatic captioning of live media streams, with a theoretical delay of 1.5 seconds. Along with the primary system, we also submitted three contrastive systems. From these, we highlight the system c2-streaming_600ms_t that, following the same configuration of the primary one, but using a smaller context window of 0.6 seconds and a Transformer LM, scored 12.3% and 16.9% WER points respectively on the same test sets, with a measured empirical latency of 0.81+-0.09 seconds (mean+-stdev). This is, we obtained state-of-the-art latencies for high-quality automatic live captioning with a small WER degradation of 6% relative. [CA] "Sistemes de reconeixement automàtic de la parla en castellà de MLLP-VRAIN per a la competició Albayzin-RTVE 2020 Speech-To-Text Challenge": En aquest article, es descriuen els sistemes de reconeixement automàtic de la parla (RAP) creats pel grup d'investigació MLLP-VRAIN de la Universitat Politecnica de València per a la competició Albayzin-RTVE 2020 Speech-to-Text Challenge. El sistema primari (p-streaming_1500ms_nlt) és un sistema de RAP híbrid BLSTM-HMM amb descodificació en temps real en una passada amb una finestra de context d'1,5 segons i una combinació lineal de models de llenguatge (ML) d'n-grames, LSTM i Transformer. El model acústic s'ha entrenat amb vora 4000 hores de parla transcrita de diferents fonts, usant el transLectures-UPV toolkit (TLK) del grup MLLP i TensorFlow; mentre que els ML s'han entrenat amb SRILM (n-grames), CUED-RNNLM (LSTM) i Fairseq (Transformer), amb 102G paraules (tokens). Aquest sistema ha obtingut 11,6 % i 16,0 % de WER en els conjunts test-2018 i test-2020, respectivament. És un sistema amb capacitat de temps real, que pot desplegar-se en producció per a subtitulació automàtica de fluxos audiovisuals en directe, amb un retard teòric d'1,5 segons. A banda del sistema primari, s'han presentat tres sistemes contrastius. D'aquests, destaquem el sistema c2-streaming_600ms_t que, amb la mateixa configuració que el sistema primari, però amb una finestra de context més reduïda de 0,6 segons i un ML Transformer, ha obtingut 12,3 % i 16,9 % de WER, respectivament, sobre els mateixos conjunts, amb una latència empírica mesurada de 0,81+-0,09 segons (mitjana+-desv). És a dir, s'han obtingut latències punteres per a subtitulació automàtica en directe d'alta qualitat amb una degradació del WER petita, del 6 % relatiu.}, keywords = {Automatic Speech Recognition, Natural Language Processing, streaming}, pubstate = {published}, tppubtype = {inproceedings} } 1st place in IberSpeech-RTVE 2020 TV Speech-to-Text Challenge. [EN] This paper describes the automatic speech recognition (ASR) systems built by the MLLP-VRAIN research group of Universitat Politecnica de València for the Albayzin-RTVE 2020 Speech-to-Text Challenge. The primary system (p-streaming_1500ms_nlt) was a hybrid BLSTM-HMM ASR system using streaming one-pass decoding with a context window of 1.5 seconds and a linear combination of an n-gram, a LSTM, and a Transformer language model (LM). The acoustic model was trained on nearly 4,000 hours of speech data from different sources, using the MLLP's transLectures-UPV toolkit (TLK) and TensorFlow; whilst LMs were trained using SRILM (n-gram), CUED-RNNLM (LSTM) and Fairseq (Transformer), with up to 102G tokens. This system achieved 11.6% and 16.0% WER on the test-2018 and test-2020 sets, respectively. As it is streaming-enabled, it could be put into production environments for automatic captioning of live media streams, with a theoretical delay of 1.5 seconds. Along with the primary system, we also submitted three contrastive systems. From these, we highlight the system c2-streaming_600ms_t that, following the same configuration of the primary one, but using a smaller context window of 0.6 seconds and a Transformer LM, scored 12.3% and 16.9% WER points respectively on the same test sets, with a measured empirical latency of 0.81+-0.09 seconds (mean+-stdev). This is, we obtained state-of-the-art latencies for high-quality automatic live captioning with a small WER degradation of 6% relative. [CA] "Sistemes de reconeixement automàtic de la parla en castellà de MLLP-VRAIN per a la competició Albayzin-RTVE 2020 Speech-To-Text Challenge": En aquest article, es descriuen els sistemes de reconeixement automàtic de la parla (RAP) creats pel grup d'investigació MLLP-VRAIN de la Universitat Politecnica de València per a la competició Albayzin-RTVE 2020 Speech-to-Text Challenge. El sistema primari (p-streaming_1500ms_nlt) és un sistema de RAP híbrid BLSTM-HMM amb descodificació en temps real en una passada amb una finestra de context d'1,5 segons i una combinació lineal de models de llenguatge (ML) d'n-grames, LSTM i Transformer. El model acústic s'ha entrenat amb vora 4000 hores de parla transcrita de diferents fonts, usant el transLectures-UPV toolkit (TLK) del grup MLLP i TensorFlow; mentre que els ML s'han entrenat amb SRILM (n-grames), CUED-RNNLM (LSTM) i Fairseq (Transformer), amb 102G paraules (tokens). Aquest sistema ha obtingut 11,6 % i 16,0 % de WER en els conjunts test-2018 i test-2020, respectivament. És un sistema amb capacitat de temps real, que pot desplegar-se en producció per a subtitulació automàtica de fluxos audiovisuals en directe, amb un retard teòric d'1,5 segons. A banda del sistema primari, s'han presentat tres sistemes contrastius. D'aquests, destaquem el sistema c2-streaming_600ms_t que, amb la mateixa configuració que el sistema primari, però amb una finestra de context més reduïda de 0,6 segons i un ML Transformer, ha obtingut 12,3 % i 16,9 % de WER, respectivament, sobre els mateixos conjunts, amb una latència empírica mesurada de 0,81+-0,09 segons (mitjana+-desv). És a dir, s'han obtingut latències punteres per a subtitulació automàtica en directe d'alta qualitat amb una degradació del WER petita, del 6 % relatiu. |
Iranzo-Sánchez, Javier; Jorge, Javier; Baquero-Arnal, Pau; Silvestre-Cerdà, Joan Albert ; Giménez, Adrià; Civera, Jorge; Sanchis, Albert; Juan, Alfons Streaming cascade-based speech translation leveraged by a direct segmentation model Journal Article Neural Networks, 142 , pp. 303–315, 2021. Abstract | Links | BibTeX | Tags: Automatic Speech Recognition, Cascade System, Deep Neural Networks, Hybrid System, Machine Translation, Segmentation Model, Speech Translation, streaming @article{Iranzo-Sánchez2021, title = {Streaming cascade-based speech translation leveraged by a direct segmentation model}, author = {Javier Iranzo-Sánchez and Javier Jorge and Pau Baquero-Arnal and Silvestre-Cerdà, Joan Albert and Adrià Giménez and Jorge Civera and Albert Sanchis and Alfons Juan}, doi = {10.1016/j.neunet.2021.05.013}, year = {2021}, date = {2021-01-01}, journal = {Neural Networks}, volume = {142}, pages = {303--315}, abstract = {The cascade approach to Speech Translation (ST) is based on a pipeline that concatenates an Automatic Speech Recognition (ASR) system followed by a Machine Translation (MT) system. Nowadays, state-of-the-art ST systems are populated with deep neural networks that are conceived to work in an offline setup in which the audio input to be translated is fully available in advance. However, a streaming setup defines a completely different picture, in which an unbounded audio input gradually becomes available and at the same time the translation needs to be generated under real-time constraints. In this work, we present a state-of-the-art streaming ST system in which neural-based models integrated in the ASR and MT components are carefully adapted in terms of their training and decoding procedures in order to run under a streaming setup. In addition, a direct segmentation model that adapts the continuous ASR output to the capacity of simultaneous MT systems trained at the sentence level is introduced to guarantee low latency while preserving the translation quality of the complete ST system. The resulting ST system is thoroughly evaluated on the real-life streaming Europarl-ST benchmark to gauge the trade-off between quality and latency for each component individually as well as for the complete ST system.}, keywords = {Automatic Speech Recognition, Cascade System, Deep Neural Networks, Hybrid System, Machine Translation, Segmentation Model, Speech Translation, streaming}, pubstate = {published}, tppubtype = {article} } The cascade approach to Speech Translation (ST) is based on a pipeline that concatenates an Automatic Speech Recognition (ASR) system followed by a Machine Translation (MT) system. Nowadays, state-of-the-art ST systems are populated with deep neural networks that are conceived to work in an offline setup in which the audio input to be translated is fully available in advance. However, a streaming setup defines a completely different picture, in which an unbounded audio input gradually becomes available and at the same time the translation needs to be generated under real-time constraints. In this work, we present a state-of-the-art streaming ST system in which neural-based models integrated in the ASR and MT components are carefully adapted in terms of their training and decoding procedures in order to run under a streaming setup. In addition, a direct segmentation model that adapts the continuous ASR output to the capacity of simultaneous MT systems trained at the sentence level is introduced to guarantee low latency while preserving the translation quality of the complete ST system. The resulting ST system is thoroughly evaluated on the real-life streaming Europarl-ST benchmark to gauge the trade-off between quality and latency for each component individually as well as for the complete ST system. |
Juan-Albarracín, Javier; Fuster-Garcia, Elies; Juan, Alfons; García-Gómez, Juan M Non-local spatially varying finite mixture models for image segmentation Journal Article Statistics and Computing, 31 (3), 2021. Abstract | Links | BibTeX | Tags: Non-local means, Spatially varying finite mixture models, Unsupervised learning @article{Juan-Albarracín2021, title = {Non-local spatially varying finite mixture models for image segmentation}, author = {Javier Juan-Albarracín and Elies Fuster-Garcia and Alfons Juan and Juan M. García-Gómez}, url = {http://hdl.handle.net/10251/183895}, doi = {10.1007/s11222-020-09988-w}, year = {2021}, date = {2021-01-01}, journal = {Statistics and Computing}, volume = {31}, number = {3}, abstract = {In this work, we propose a new Bayesian model for unsupervised image segmentation based on a combination of the spatially varying finite mixture models (SVFMMs) and the non-local means (NLM) framework. The probabilistic NLM weighting function is successfully integrated into a varying Gauss–Markov random field, yielding a prior density that adaptively imposes a local regularization to simultaneously preserve edges and enforce smooth constraints in homogeneous regions of the image. Two versions of our model are proposed: a pixel-based model and a patch-based model, depending on the design of the probabilistic NLM weighting function. Contrary to previous methods proposed in the literature, our approximation does not introduce new parameters to be estimated into the model, because the NLM weighting function is completely known once the neighborhood of a pixel is fixed. The proposed model can be estimated in closed-form solution via a maximum a posteriori (MAP) estimation in an expectation–maximization scheme. We have compared our model with previously proposed SVFMMs using two public datasets: the Berkeley Segmentation dataset and the BRATS 2013 dataset. The proposed model performs favorably to previous approaches in the literature, achieving better results in terms of Rand Index and Dice metrics in our experiments.}, keywords = {Non-local means, Spatially varying finite mixture models, Unsupervised learning}, pubstate = {published}, tppubtype = {article} } In this work, we propose a new Bayesian model for unsupervised image segmentation based on a combination of the spatially varying finite mixture models (SVFMMs) and the non-local means (NLM) framework. The probabilistic NLM weighting function is successfully integrated into a varying Gauss–Markov random field, yielding a prior density that adaptively imposes a local regularization to simultaneously preserve edges and enforce smooth constraints in homogeneous regions of the image. Two versions of our model are proposed: a pixel-based model and a patch-based model, depending on the design of the probabilistic NLM weighting function. Contrary to previous methods proposed in the literature, our approximation does not introduce new parameters to be estimated into the model, because the NLM weighting function is completely known once the neighborhood of a pixel is fixed. The proposed model can be estimated in closed-form solution via a maximum a posteriori (MAP) estimation in an expectation–maximization scheme. We have compared our model with previously proposed SVFMMs using two public datasets: the Berkeley Segmentation dataset and the BRATS 2013 dataset. The proposed model performs favorably to previous approaches in the literature, achieving better results in terms of Rand Index and Dice metrics in our experiments. |
Garcés Díaz-Munío, Gonçal V; Silvestre-Cerdà, Joan Albert ; Jorge, Javier; Giménez, Adrià; Iranzo-Sánchez, Javier; Baquero-Arnal, Pau; Roselló, Nahuel; Pérez-González-de-Martos, Alejandro; Civera, Jorge; Sanchis, Albert; Juan, Alfons Europarl-ASR: A Large Corpus of Parliamentary Debates for Streaming ASR Benchmarking and Speech Data Filtering/Verbatimization Inproceedings Proc. Interspeech 2021, pp. 3695–3699, Brno (Czech Republic), 2021. Abstract | Links | BibTeX | Tags: Automatic Speech Recognition, speech corpus, speech data filtering, speech data verbatimization @inproceedings{Garcés2021, title = {Europarl-ASR: A Large Corpus of Parliamentary Debates for Streaming ASR Benchmarking and Speech Data Filtering/Verbatimization}, author = {Garcés Díaz-Munío, Gonçal V. and Silvestre-Cerdà, Joan Albert and Javier Jorge and Adrià Giménez and Javier Iranzo-Sánchez and Pau Baquero-Arnal and Nahuel Roselló and Alejandro Pérez-González-de-Martos and Jorge Civera and Albert Sanchis and Alfons Juan}, url = {https://www.mllp.upv.es/wp-content/uploads/2021/09/europarl-asr-presentation-extended.pdf https://www.youtube.com/watch?v=Tc0gNSDdnQg&list=PLlePn-Yanvnc_LRhgmmaNmH12Bwm6BRsZ}, doi = {10.21437/Interspeech.2021-1905}, year = {2021}, date = {2021-01-01}, booktitle = {Proc. Interspeech 2021}, journal = {Proc. Interspeech 2021}, pages = {3695--3699}, address = {Brno (Czech Republic)}, abstract = {[EN] We introduce Europarl-ASR, a large speech and text corpus of parliamentary debates including 1300 hours of transcribed speeches and 70 million tokens of text in English extracted from European Parliament sessions. The training set is labelled with the Parliament’s non-fully-verbatim official transcripts, time-aligned. As verbatimness is critical for acoustic model training, we also provide automatically noise-filtered and automatically verbatimized transcripts of all speeches based on speech data filtering and verbatimization techniques. Additionally, 18 hours of transcribed speeches were manually verbatimized to build reliable speaker-dependent and speaker-independent development/test sets for streaming ASR benchmarking. The availability of manual non-verbatim and verbatim transcripts for dev/test speeches makes this corpus useful for the assessment of automatic filtering and verbatimization techniques. This paper describes the corpus and its creation, and provides off-line and streaming ASR baselines for both the speaker-dependent and speaker-independent tasks using the three training transcription sets. The corpus is publicly released under an open licence. [CA] "Europarl-ASR: Un extens corpus parlamentari de referència per a reconeixement de la parla i filtratge/literalització de transcripcions": Presentem Europarl-ASR, un extens corpus de veu i text de debats parlamentaris amb 1300 hores d'intervencions transcrites i 70 milions de paraules de text en anglés extrets de sessions del Parlament Europeu. Les transcripcions oficials del Parlament Europeu, no literals, s'han sincronitzat per a tot el conjunt d'entrenament. Com que l'entrenament de models acústics requereix transcripcions com més literals millor, també s'han inclòs transcripcions filtrades i transcripcions literalitzades de totes les intervencions, basades en tècniques de filtratge i literalització automàtics. A més, s'han inclòs 18 hores de transcripcions literals revisades manualment per definir dos conjunts de validació i avaluació de referència per a reconeixement automàtic de la parla en temps real, amb oradors coneguts i amb oradors desconeguts. Pel fet de disposar de transcripcions literals i no literals, aquest corpus és també ideal per a l'anàlisi de tècniques de filtratge i de literalització. En aquest article, es descriu la creació del corpus i es proporcionen mesures de referència de reconeixement automàtic de la parla en temps real i en diferit, amb oradors coneguts i amb oradors desconeguts, usant els tres conjunts de transcripcions d'entrenament. El corpus es fa públic amb una llicència oberta.}, keywords = {Automatic Speech Recognition, speech corpus, speech data filtering, speech data verbatimization}, pubstate = {published}, tppubtype = {inproceedings} } [EN] We introduce Europarl-ASR, a large speech and text corpus of parliamentary debates including 1300 hours of transcribed speeches and 70 million tokens of text in English extracted from European Parliament sessions. The training set is labelled with the Parliament’s non-fully-verbatim official transcripts, time-aligned. As verbatimness is critical for acoustic model training, we also provide automatically noise-filtered and automatically verbatimized transcripts of all speeches based on speech data filtering and verbatimization techniques. Additionally, 18 hours of transcribed speeches were manually verbatimized to build reliable speaker-dependent and speaker-independent development/test sets for streaming ASR benchmarking. The availability of manual non-verbatim and verbatim transcripts for dev/test speeches makes this corpus useful for the assessment of automatic filtering and verbatimization techniques. This paper describes the corpus and its creation, and provides off-line and streaming ASR baselines for both the speaker-dependent and speaker-independent tasks using the three training transcription sets. The corpus is publicly released under an open licence. [CA] "Europarl-ASR: Un extens corpus parlamentari de referència per a reconeixement de la parla i filtratge/literalització de transcripcions": Presentem Europarl-ASR, un extens corpus de veu i text de debats parlamentaris amb 1300 hores d'intervencions transcrites i 70 milions de paraules de text en anglés extrets de sessions del Parlament Europeu. Les transcripcions oficials del Parlament Europeu, no literals, s'han sincronitzat per a tot el conjunt d'entrenament. Com que l'entrenament de models acústics requereix transcripcions com més literals millor, també s'han inclòs transcripcions filtrades i transcripcions literalitzades de totes les intervencions, basades en tècniques de filtratge i literalització automàtics. A més, s'han inclòs 18 hores de transcripcions literals revisades manualment per definir dos conjunts de validació i avaluació de referència per a reconeixement automàtic de la parla en temps real, amb oradors coneguts i amb oradors desconeguts. Pel fet de disposar de transcripcions literals i no literals, aquest corpus és també ideal per a l'anàlisi de tècniques de filtratge i de literalització. En aquest article, es descriu la creació del corpus i es proporcionen mesures de referència de reconeixement automàtic de la parla en temps real i en diferit, amb oradors coneguts i amb oradors desconeguts, usant els tres conjunts de transcripcions d'entrenament. El corpus es fa públic amb una llicència oberta. |
Pérez-González-de-Martos, Alejandro; Iranzo-Sánchez, Javier; Giménez Pastor, Adrià ; Jorge, Javier; Silvestre-Cerdà, Joan-Albert; Civera, Jorge; Sanchis, Albert; Juan, Alfons Towards simultaneous machine interpretation Inproceedings Proc. Interspeech 2021, pp. 2277–2281, Brno (Czech Republic), 2021. Abstract | Links | BibTeX | Tags: cross-lingual voice cloning, incremental text-to-speech, simultaneous machine interpretation, speech-to-speech translation @inproceedings{Pérez-González-de-Martos2021, title = {Towards simultaneous machine interpretation}, author = {Alejandro Pérez-González-de-Martos and Javier Iranzo-Sánchez and Giménez Pastor, Adrià and Javier Jorge and Joan-Albert Silvestre-Cerdà and Jorge Civera and Albert Sanchis and Alfons Juan}, doi = {10.21437/Interspeech.2021-201}, year = {2021}, date = {2021-01-01}, booktitle = {Proc. Interspeech 2021}, journal = {Proc. Interspeech 2021}, pages = {2277--2281}, address = {Brno (Czech Republic)}, abstract = {Automatic speech-to-speech translation (S2S) is one of the most challenging speech and language processing tasks, especially when considering its application to real-time settings. Recent advances in streaming Automatic Speech Recognition (ASR), simultaneous Machine Translation (MT) and incremental neural Text-To-Speech (TTS) make it possible to develop real-time cascade S2S systems with greatly improved accuracy. On the way to simultaneous machine interpretation, a state-of-the-art cascade streaming S2S system is described and empirically assessed in the simultaneous interpretation of European Parliament debates. We pay particular attention to the TTS component, particularly in terms of speech naturalness under a variety of response-time settings, as well as in terms of speaker similarity for its cross-lingual voice cloning capabilities.}, keywords = {cross-lingual voice cloning, incremental text-to-speech, simultaneous machine interpretation, speech-to-speech translation}, pubstate = {published}, tppubtype = {inproceedings} } Automatic speech-to-speech translation (S2S) is one of the most challenging speech and language processing tasks, especially when considering its application to real-time settings. Recent advances in streaming Automatic Speech Recognition (ASR), simultaneous Machine Translation (MT) and incremental neural Text-To-Speech (TTS) make it possible to develop real-time cascade S2S systems with greatly improved accuracy. On the way to simultaneous machine interpretation, a state-of-the-art cascade streaming S2S system is described and empirically assessed in the simultaneous interpretation of European Parliament debates. We pay particular attention to the TTS component, particularly in terms of speech naturalness under a variety of response-time settings, as well as in terms of speaker similarity for its cross-lingual voice cloning capabilities. |
Javier Iranzo-Sánchez Jorge Civera, Alfons Juan Stream-level Latency Evaluation for Simultaneous Machine Translation Inproceedings Findings of the ACL: EMNLP 2021, pp. 664–670, Punta Cana (Dominican Republic), 2021. Abstract | Links | BibTeX | Tags: latency, simultaneous machine translation, stream-level evaluation, streaming @inproceedings{Iranzo-Sánchez2021b, title = {Stream-level Latency Evaluation for Simultaneous Machine Translation}, author = {Javier Iranzo-Sánchez, Jorge Civera, Alfons Juan}, url = {https://arxiv.org/abs/2104.08817 https://github.com/jairsan/Stream-level_Latency_Evaluation_for_Simultaneous_Machine_Translation}, doi = {10.18653/v1/2021.findings-emnlp.58}, year = {2021}, date = {2021-01-01}, booktitle = {Findings of the ACL: EMNLP 2021}, pages = {664--670}, address = {Punta Cana (Dominican Republic)}, abstract = {Simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications. Simultaneous translation systems need to find a trade-off between translation quality and response time, and with this purpose multiple latency measures have been proposed. However, latency evaluations for simultaneous translation are estimated at the sentence level, not taking into account the sequential nature of a streaming scenario. Indeed, these sentence-level latency measures are not well suited for continuous stream translation, resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed. This work proposes a stream level adaptation of the current latency measures based on a re-segmentation approach applied to the output translation, that is successfully evaluated on streaming conditions for a reference IWSLT task.}, keywords = {latency, simultaneous machine translation, stream-level evaluation, streaming}, pubstate = {published}, tppubtype = {inproceedings} } Simultaneous machine translation has recently gained traction thanks to significant quality improvements and the advent of streaming applications. Simultaneous translation systems need to find a trade-off between translation quality and response time, and with this purpose multiple latency measures have been proposed. However, latency evaluations for simultaneous translation are estimated at the sentence level, not taking into account the sequential nature of a streaming scenario. Indeed, these sentence-level latency measures are not well suited for continuous stream translation, resulting in figures that are not coherent with the simultaneous translation policy of the system being assessed. This work proposes a stream level adaptation of the current latency measures based on a re-segmentation approach applied to the output translation, that is successfully evaluated on streaming conditions for a reference IWSLT task. |
Pérez-González-de-Martos, Alejandro; Sanchis, Albert; Juan, Alfons VRAIN-UPV MLLP's system for the Blizzard Challenge 2021 Inproceedings Proc. of Blizzard Challenge 2021, 2021. Abstract | Links | BibTeX | Tags: Blizzard Challenge, HiFi-GAN, text-to-speech @inproceedings{Pérez-González-de-Martos2021b, title = {VRAIN-UPV MLLP's system for the Blizzard Challenge 2021}, author = {Alejandro Pérez-González-de-Martos and Albert Sanchis and Alfons Juan}, url = {http://hdl.handle.net/10251/192554 https://arxiv.org/abs/2110.15792 http://www.festvox.org/blizzard/blizzard2021.html}, year = {2021}, date = {2021-01-01}, booktitle = {Proc. of Blizzard Challenge 2021}, abstract = {This paper presents the VRAIN-UPV MLLP’s speech synthesis system for the SH1 task of the Blizzard Challenge 2021. The SH1 task consisted in building a Spanish text-to-speech system trained on (but not limited to) the corpus released by the Blizzard Challenge 2021 organization. It included 5 hours of studio-quality recordings from a native Spanish female speaker. In our case, this dataset was solely used to build a two-stage neural text-to-speech pipeline composed of a non-autoregressive acoustic model with explicit duration modeling and a HiFi-GAN neural vocoder. Our team is identified as J in the evaluation results. Our system obtained very good results in the subjective evaluation tests. Only one system among other 11 participants achieved better naturalness than ours. Concretely, it achieved a naturalness MOS of 3.61 compared to 4.21 for real samples.}, keywords = {Blizzard Challenge, HiFi-GAN, text-to-speech}, pubstate = {published}, tppubtype = {inproceedings} } This paper presents the VRAIN-UPV MLLP’s speech synthesis system for the SH1 task of the Blizzard Challenge 2021. The SH1 task consisted in building a Spanish text-to-speech system trained on (but not limited to) the corpus released by the Blizzard Challenge 2021 organization. It included 5 hours of studio-quality recordings from a native Spanish female speaker. In our case, this dataset was solely used to build a two-stage neural text-to-speech pipeline composed of a non-autoregressive acoustic model with explicit duration modeling and a HiFi-GAN neural vocoder. Our team is identified as J in the evaluation results. Our system obtained very good results in the subjective evaluation tests. Only one system among other 11 participants achieved better naturalness than ours. Concretely, it achieved a naturalness MOS of 3.61 compared to 4.21 for real samples. |