检索结果-内蒙古大学图书馆

Incremental Lattice Determinization for WFST Decoders

学校读者我要写书评

暂无评论

Incremental Lattice Determinization for WFST Decoders

IEEE Workshop on Automatic speech Recognition and Understanding

作者： Zhehuai Chen Mahsa Yarmohammadi Hainan Xu Hang Lv Lei Xie Daniel Povey Sanjeev Khudanpur SpeechLab Shanghai Jiao Tong University Center for Language and Speech Processing ASLP@NPU School of Computer Science Northwestern Polytechnical University HLTCOE Johns Hopkins University

ISBN: (数字)9781728103068

ISBN: (纸本)9781728103075

We introduce a lattice determinization algorithm that can operate incrementally. That is, a word-level lattice can be generated for a partial utterance and then, once we have processed more audio, we can obtain a word-level lattice for the extended utterance without redoing all the work of lattice determinization. This is relevant for ASR decoders such as those used in Kaldi, which first generate a state-level lattice and then convert it to a word-level lattice using a determinization algorithm in a special semiring. Our incremental determinization algorithm is useful when word-level lattices are needed prior to the end of the utterance, and also reduces the latency due to determinization at the end of the utterance.

关键词： Lattices Decoding Real-time systems Automata Transducers Task analysis computer science

MultiplEYE: Creating a multilingual eye-tracking-while-reading corpus 25

学校读者我要写书评

暂无评论

MultiplEYE: Creating a multilingual eye-tracking-while-readi...

Proceedings of the 2025 Symposium on Eye Tracking Research and Applications

作者： Deborah Noemie Jakobi Maja Stegenwallner-Schütz Nora Hollenstein Cui Ding Ramune Kaspere Ana Matić Škorić Eva Pavlinusic Vilus Stefan Frank Marie-Luise Müller Kristine M Jensen de López Nik Kharlamov Hanne B. Søndergaard Knudsen Yevgeni Berzak Ella Lion Irina A. Sekerina Cengiz Acarturk Mohd Faizan Ansari Katarzyna Harezlak Pawel Kasprowski Ana Bautista Lisa Beinborn Anna Bondar Antonia Boznou Leah Bradshaw Jana Mara Hofmann Thyra Krosness Not Battesta Soliva Anila Çepani Kristina Cergol Ana Došen Marijan Palmovic Adelina Çerpja Dalí Chirino Jan Chromý Vera Demberg Iza Škrjanec Nazik Dinçtopal Deniz Dr. Inmaculada Fajardo Mariola Giménez-Salvador Xavier Mínguez-López Maroš Filip Zigmunds Freibergs Jéssica Gomes Andreia Janeiro Paula Luegi João Veríssimo Sasho Gramatikov Jana Hasenäcker Alba Haveriku Nelda Kote Muhammad M. Kamal Hanna Kędzierska Dorota Klimek-Jankowska Sara Kosutar Daniel G. Krakowczyk Izabela Krejtz Marta Łockiewicz Kaidi Lõo Jurgita Motiejūnienė Jamal A. Nasir Johanne Sofie Krog Nedergård Ayşegül Özkan Mikuláš Preininger Loredana Pungă David Robert Reich Chiara Tschirner Špela Rot Andreas Säuberli Jordi Solé-Casals Ekaterina Strati Igor Svoboda Evis Trandafili Spyridoula Varlokosta Mila Vulchanova Lena A. Jäger Department of Computational Linguistics University of Zurich Zurich Switzerland Humboldt-University of Berlin Berlin Germany and University of Koblenz Koblenz Germany Kaunas University of Technology Kaunas Lithuania Faculty of Education and Rehabilitation Sciences Dept of Speech and Language pathology Laboratory for Psycholinguistic Research University of Zagreb Zagreb Croatia Centre for Language Studies Radboud University Nijmegen Netherlands ZPID Leibnitz Institute for Psychology Trier Germany Aalborg University Aalborg Denmark Data and Decision Sciences Technion Haifa Israel Department of Psychology College of Staten Island Staten Island New York USA Cognitive Science Department Jagiellonian University Krakow Poland Silesian University of Technology Gliwice Poland Institute of Informatics Silesian University of Technology Gliwice Poland Basque Center on Cognition Brain & Language Donostia Spain Human-Centered Data Science University of Göttingen Göttingen Germany Department of Computational Linguistics University of Zurich Zurich Switzerland and Digital Society Initiative University of Zurich Zurich Switzerland Lab of Psycholinguistics & Neurolinguistics National and Kapodistrian University of Athens Athens Greece Department of Computational Linguistics University of Zurich University of Zurich Switzerland University of Zurich Zurich Switzerland University of Tirana Tirana Albania Faculty of Teacher Education University of Zagreb Zagreb Croatia Hrvatska Croatia and Laboratory for Psycholinguistic Research University of Zagreb Zagreb Croatia Hrvatska Croatia University of Zagreb Zagreb Croatia Dept. of Speech & Lang. Path. Laboratory for Psycholinguistic Research University of Zagreb Zagreb Croatia Hrvatska Croatia Academy of Sciences of Albania Tirana Albania Radboud University Nijmegen Netherlands Institute of Czech Language and Theory of Communication Faculty of Arts Charles University Prague Czech Republic Saarland Universit

Improving Parallel Corpus Quality for Chinese-Vietnamese Statistical Machine Translation

学校读者我要写书评

暂无评论

Journal of Beijing Institute of Technology 2018年第1期27卷 127-136页

作者： Huu-anh Tran Yuhang Guo Ping Jian Shumin Shi Heyan Huang Department of Computer Science and Technology Beijing Institute of Technology Beijing Engineering Research Center of High Volume Language Information Processing and Cloud Computing Application Beijing Institute of Technology

The performance of a machine translation system heavily depends on the quantity and quality of the bilingual language resource. However,getting a parallel corpus,which has a large scale and is of high quality,is a very difficult task especially for low resource languages such as Chinese-Vietnamese. Fortunately,multilingual user generated contents（ UGC）,such as bilingual movie subtitles,provide us access to automatic construction of the parallel corpus. Although the amount of UGC parallel corpora can be considerable,the original corpus is not suitable for statistical machine translation（ SMT） systems. The corpus may contain translation errors,sentence mismatching,free translations,etc. To improve the quality of the bilingual corpus for SMT systems,three filtering methods are proposed： sentence length difference,the semantic of sentence pairs,and machine learning. Experiments are conducted on the Chinese to Vietnamese translation *** results demonstrate that all the three methods effectively improve the corpus quality,and the machine translation performance（ BLEU score） can be improved by 1. 32.

关键词： parallel corpus filtering low resource languages bilingual movie subtitles machine translation Chinese-Vietnamese translation

Using ASR Methods for OCR

学校读者我要写书评

暂无评论

Using ASR Methods for OCR

International Conference on Document Analysis and Recognition

作者： Ashish Arora Chun Chieh Chang Babak Rekabdar Bagher BabaAli Daniel Povey David Etter Desh Raj Hossein Hadian Jan Trmal Paola Garcia Shinji Watanabe Vimal Manohar Yiwen Shao Sanjeev Khudanpur Center for Language and Speech Processing Johns Hopkins University Baltimore USA Human Language Technology Center of Excellence Johns Hopkins University Baltimore USA School of Mathematics Statistics and Computer Sciences University of Tehran Iran Department of Computer Engineering Sharif University of Technology Iran

Hybrid deep neural network hidden Markov models (DNN-HMM) have achieved impressive results on large vocabulary continuous speech recognition (LVCSR) tasks. However, the recent approaches using DNN-HMM models are not explored much for text recognition. Inspired by the current work in automatic speech recognition (ASR) and machine translation, we present an open vocabulary sub-word text recognition system. The sub-word lexicon and sub-word language model (LM) helps in overcoming the challenge of recognizing out of vocabulary (OOV) words, and a time delay neural network (TDNN) and convolution neural network (CNN) based DNN-HMM optical model (OM) efficiently models the sequence dependency in the line image. We present results on 12 datasets with training data varying from 6k lines to 600k lines. The system is built for 8 languages, i.e., English, French, Arabic, Chinese, Farsi, Tamil, Russian, and Korean. We report competitive results on several commonly used handwritten and printed text datasets.

关键词： Hidden Markov models Text recognition Vocabulary Training Optical imaging Training data Feature extraction

CAA-net: Conditional atrous CNNs with attention for explainable device-robust acoustic scene classification

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Ren, Zhao Kong, Qiuqiang Han, Jing Plumbley, Mark D. Schuller, Björn W. Chair of Embedded Intelligence for Health Care and Wellbeing University of Augsburg Germany Centre for Vision Speech and Signal Processing University of Surrey United Kingdom Department of Computer Science and Technology University of Cambridge United Kingdom GLAM – Group on Language Audio & Music Imperial College London United Kingdom Chair of Embedded Intelligence for Health Care and Wellbeing University of Augsburg Germany

Acoustic Scene Classification (ASC) aims to classify the environment in which the audio signals are recorded. Recently, Convolutional Neural Networks (CNNs) have been successfully applied to ASC. However, the data distributions of the audio signals recorded with multiple devices are different. There has been little research on the training of robust neural networks on acoustic scene datasets recorded with multiple devices, and on explaining the operation of the internal layers of the neural networks. In this article, we focus on training and explaining device-robust CNNs on multi-device acoustic scene data. We propose conditional atrous CNNs with attention for multi-device ASC. Our proposed system contains an ASC branch and a device classification branch, both modelled by CNNs. We visualise and analyse the intermediate layers of the atrous CNNs. A time-frequency attention mechanism is employed to analyse the contribution of each time-frequency bin of the feature maps in the CNNs. On the Detection and Classification of Acoustic Scenes and Events (DCASE) 2018 ASC dataset, recorded with three devices, our proposed model performs significantly better than CNNs trained on single-device data. Copyright © 2020, The Authors. All rights reserved.

关键词： Convolutional neural networks

On the impact of various types of noise on neural machine translation

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Khayrallah, Huda Koehn, Philipp Center for Language & Speech Processing Computer Science Department Johns Hopkins University

We examine how various types of noise in the parallel training data impact the quality of neural machine translation systems. We create five types of artificial noise and analyze how they degrade performance in neural and statistical machine translation. We find that neural models are generally more harmed by noise than statistical models. For one especially egregious type of noise they learn to just copy the input sentence. Copyright © 2018, The Authors. All rights reserved.

关键词： Neural machine translation

MCE 2018: The 1st multi-target speaker detection and identification challenge evaluation

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Shon, Suwon Dehak, Najim Reynolds, Douglas Glass, James MIT Computer Science and Artificial Intelligence Laboratory CambridgeMA United States Center for Language and Speech Processing Johns Hopkins University Baltimore United States MIT Lincoln Laboratory LexingtonMA United States

The Multi-target Challenge1 aims to assess how well current speech technology is able to determine whether or not a recorded utterance was spoken by one of a large number of blacklisted speakers. It is a form of multi-target speaker detection based on real-world telephone conversations. Data recordings are generated from call center customer-agent conversations. The task is to measure how accurately one can detect 1) whether a test recording is spoken by a blacklisted speaker, and 2) which specific blacklisted speaker was talking. This paper outlines the challenge and provides its baselines, results, and discussions. Copyright © 2019, The Authors. All rights reserved.

关键词： speech recognition

End-to-end speech recognition using lattice-free MMI 19

学校读者我要写书评

暂无评论

End-to-end speech recognition using lattice-free MMI

19th Annual Conference of the International speech Communication, INTERspeech 2018

作者： Hadian, Hossein Sameti, Hossein Povey, Daniel Khudanpur, Sanjeev Department of Computer Engineering Sharif University of Technology Tehran Iran Center for Language and Speech Processing Johns Hopkins University BaltimoreMD United States Human Language Technology Center of Excellence Johns Hopkins University Baltimore United States

We present our work on end-to-end training of acoustic models using the lattice-free maximum mutual information (LF-MMI) objective function in the context of hidden Markov models. By end-to-end training, we mean flat-start training of a single DNN in one stage without using any previously trained models, forced alignments, or building state-tying decision trees. We use full biphones to enable context-dependent modeling without trees, and show that our end-to-end LF-MMI approach can achieve comparable results to regular LF-MMI on well-known large vocabulary tasks. We also compare with other end-to-end methods such as CTC in character-based and lexicon-free settings and show 5 to 25 percent relative reduction in word error rates on different large vocabulary tasks while using significantly smaller models. © 2018 International speech Communication Association. All rights reserved.

关键词： Hidden Markov models

Promoting diversity for end-to-end conversation response generation

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Ruan, Yu-Ping Ling, Zhen-Hua Liu, Quan Gu, Jia-Chen Zhu, Xiaodan National Engineering Laboratory for Speech and Language Information Processing University of Science and Technology of China Hefei China iFLYTEK Research Hefei China Department of Electrical and Computer Engineering Queen's University Kingston Canada

We present our work on Track 2 in the Dialog System Technology Challenges 7 (DSTC7). The DSTC7-Track 2 aims to evaluate the response generation of fully data-driven conversation models in knowledge-grounded settings, which provides the contextual-relevant factual texts. The Sequence-to-Sequence models have been widely used for end-to-end generative conversation modelling and achieved impressive results. However, they tend to output dull and repeated responses in previous studies. Our work aims to promote the diversity for end-to-end conversation response generation, which follows a two-stage pipeline: 1) Generate multiple responses. At this stage, two different models are proposed, i.e., a variational generative (VariGen) model and a retrieval based (Retrieval) model. 2) Rank and return the most related response by training a topic coherence discrimination (TCD) model for the ranking process. According to the official evaluation results, our proposed Retrieval and VariGen systems ranked first and second respectively on objective diversity metrics, i.e., Entropy, among all participant systems. And the VariGen system ranked second on NIST and METEOR metrics. Copyright © 2019, The Authors. All rights reserved.

关键词：