Berthing and Priority Policies and Procedures A - Berthing Priority Policy 1. All commercial vessels utilising TasPorts ports are scheduled through TasPorts’ VTS Centre at Bell Bay; 2. All vessel arrivals will submit a

Source URL: www.tasports.com.au

SEQUENCE-TO-SEQUENCE MODELS FOR PUNCTUATED TRANSCRIPTION COMBINING LEXICAL AND ACOUSTIC FEATURES Ondˇrej Klejch, Peter Bell, Steve Renals Centre for Speech Technology Research, University of Edinburgh, Edinburgh EH8 9AB

Source URL: www.cstr.ed.ac.uk

REGULARIZATION OF CONTEXT-DEPENDENT DEEP NEURAL NETWORKS WITH CONTEXT-INDEPENDENT MULTI-TASK TRAINING Peter Bell and Steve Renals Centre for Speech Technology Research, University of Edinburgh, Edinburgh EH8 9AB, UK ABST

Source URL: www.cstr.inf.ed.ac.uk

A system for automatic broadcast news summarisation, geolocation and translation Peter Bell, Catherine Lai, Clare Llewellyn, Alexandra Birch, Mark Sinclair Centre for Speech Technology Research, University of Edinburgh,

Source URL: www.cstr.inf.ed.ac.uk

• Computational linguistics
• Speech recognition
• Natural language processing
• Statistical machine translation
• Machine translation
• Named-entity recognition
• Text segmentation
• Search engine indexing
• Information extraction

Sustainability Governance and Indicators David V. J Bell Director, York Centre for Applied

Source URL: www.sustainabilityindicators.org

• Accountability
• Governance
• Political engineering
• Political philosophy
• Sustainability
• Sustainable forest management
• Sustainability metrics and indices
• Melbourne Principles

Complementary tasks for context-dependent deep neural network acoustic models Peter Bell, Steve Renals Centre for Speech Technology Research, University of Edinburgh, Edinburgh EH8 9AB, UK {peter.bell, s.renals}@ed.ac.uk

Source URL: homepages.inf.ed.ac.uk

• Machine learning
• Computational neuroscience
• Computational linguistics
• Computational statistics
• Computer accessibility
• Speech recognition
• Artificial neural network
• Multi-task learning
• Recurrent neural network
• Statistical classification
• Deep learning

Complementary tasks for context-dependent deep neural network acoustic models Peter Bell, Steve Renals Centre for Speech Technology Research, University of Edinburgh, Edinburgh EH8 9AB, UK {peter.bell, s.renals}@ed.ac.uk

Source URL: www.cstr.inf.ed.ac.uk

• Machine learning
• Computational linguistics
• Computational neuroscience
• Automatic identification and data capture
• Computer accessibility
• Speech recognition
• Artificial neural network
• Multi-task learning
• Recurrent neural network
• Statistical classification
• Deep learning

Structured Output Layer with Auxiliary Targets for Context-Dependent Acoustic Modelling Pawel Swietojanski, Peter Bell, Steve Renals The Centre for Speech Technology Research, University of Edinburgh, United Kingdom {p.s

Source URL: www.cstr.inf.ed.ac.uk

• Artificial neural networks
• Computational neuroscience
• Computational statistics
• Automatic identification and data capture
• Computer accessibility
• Speech recognition
• Language model
• Convolutional neural network
• Perceptron
• Recurrent neural network
• Machine learning
• Deep learning

PUBLIC INTEREST ADVOCACY CENTRE LE CENTRE POUR LA DÉFENSE DE L’INTÉRÊT PUBLIC ONE Nicholas Street, Suite 1204, Ottawa, Ontario, Canada K1N 7B7 Tel: (Fax: (Website: http://www.piac.ca 1

Source URL: www.piac.ca

• Quebecor
• Vidotron
• Bell Canada
• Canadian Radio-television and Telecommunications Commission
• Bell Mobility
• Globe

A SYSTEM FOR AUTOMATIC ALIGNMENT OF BROADCAST MEDIA CAPTIONS USING WEIGHTED FINITE-STATE TRANSDUCERS Peter Bell and Steve Renals Centre for Speech Technology Research, University of Edinburgh, Edinburgh EH8 9AB, UK {pete

Source URL: homepages.inf.ed.ac.uk

• Computational linguistics
• Speech recognition
• Acoustic model
• Alignment
• Speech synthesis
• Closed captioning
• Word error rate
• Word-sense disambiguation
• Language model
• Finite-state transducer