CALO Decoder Progress Report for April/May

Slides:



Advertisements
Similar presentations
Speech Recognition with Hidden Markov Models Winter 2011
Advertisements

CALO Decoder Progress Report for March Arthur (Decoder and ICSI Training) Jahanzeb (Decoder) Ziad (ICSI Training) Moss (ICSI Training) Carnegie Mellon.
HIWIRE MEETING Paris, February 11, 2005 JOSÉ C. SEGURA LUNA GSTC UGR.
Teresa K. Goetze SharePoint Experience, Software Skills and Site Samples.
Brief Overview of Different Versions of Sphinx Arthur Chan.
Progress of Sphinx 3.X From X=5 to X=6 Arthur Chan Evandro Gouvea David J. Huggins-Daines Alex I. Rudnicky Mosur Ravishankar Yitao Sun.
CALO Recorder/Decoder Progress Report for Summer 2004 (July and August) Yitao Sun (Recorder/Decoder) Jason Cohen (Recorder/End-pointer) Thomas Quisel (Recorder)
Speaker Adaptation in Sphinx 3.x and CALO David Huggins-Daines
Extending VERA (Conference Information) Design Specification & Schedules Arthur Chan (AC) Rohit Kumar (RK) Lingyun Gao (LG)
Progress of Sphinx 3.X, From X=4 to X=5 By Arthur Chan Evandro Gouvea Yitao Sun David Huggins-Daines Jahanzeb Sherwani.
Almost-Spring Short Course on Speech Recognition Instructors: Bhiksha Raj and Rita Singh Welcome.
Technical Aspects of the CALO Recorder By Satanjeev Banerjee Thomas Quisel Jason Cohen Arthur Chan Yitao Sun David Huggins-Daines Alex Rudnicky.
1 USING CLASS WEIGHTING IN INTER-CLASS MLLR Sam-Joo Doh and Richard M. Stern Department of Electrical and Computer Engineering and School of Computer Science.
Part 2: Requirements Days 7, 9, 11, 13 Chapter 2: How to Gather Requirements: Some Techniques to Use Chapter 3: Finding Out about the Users and the Domain.
Sphinx 3.4 Development Progress Arthur Chan, Jahanzeb Sherwani Carnegie Mellon University Mar 4, 2004.
CALO Decoder Progress Report for June Arthur (Decoder, Trainer, ICSI Training) Yitao (Live-mode Decoder) Ziad (ICSI Training) Carnegie Mellon University.
Sphinx 3.4 Development Progress Report in February Arthur Chan, Jahanzeb Sherwani Carnegie Mellon University Mar 1, 2004.
1 Mouse Cases M.Ellis/Y.Torun Wednesday 31 st March 2004.
Unit 2 Present Progressive and Simple Present. Unit 2 Present Progressive and Simple Present 2 Present Progressive.
Adaptation Techniques in Automatic Speech Recognition Tor André Myrvoll Telektronikk 99(2), Issue on Spoken Language Technology in Telecommunications,
Bryce Rodgers Kent Warner Matt Heckman.
1M4 speech recognition University of Sheffield M4 speech recognition Martin Karafiát*, Steve Renals, Vincent Wan.
1 Design and Performance of a Web Server Accelerator Eric Levy-Abegnoli, Arun Iyengar, Junehwa Song, and Daniel Dias INFOCOM ‘99.
Publish Calendars to the Web. CCUweb Presentation (10 Minutes) 1 Demonstration of published calendars (10 minutes) 2 Demonstration of importing calendar.
Max Planck Institute for Psycholinguistics Tool development report H. Brugman MPI Nijmegen.
Project Tracking. Questions... Why should we track a project that is underway? What aspects of a project need tracking?
Nightly Releases and Testing Alexander Undrus Atlas SW week, May
CMU Shpinx Speech Recognition Engine Reporter : Chun-Feng Liao NCCU Dept. of Computer Sceince Intelligent Media Lab.
Database-Driven Web Sites, Second Edition1 Chapter 5 WEB SERVERS.
Comparison of the SPHINX and HTK Frameworks Processing the AN4 Corpus Arthur Kunkle ECE 5526 Fall 2008.
Speaker Diarisation and Large Vocabulary Recognition at CSTR: The AMI/AMIDA System Fergus McInnes 7 December 2011 History – AMI, AMIDA and recent developments.
1 Improved Speaker Adaptation Using Speaker Dependent Feature Projections Spyros Matsoukas and Richard Schwartz Sep. 5, 2003 Martigny, Switzerland.
Chapter 25: Code-Tuning Strategies. Chapter 25  Code tuning is one way of improving a program’s performance, You can often find other ways to improve.
Porto, 4-5 March, 1999 The COST250 Speaker Recognition Reference System H. Melin, A.M. Ariyaeeinia, M. Falcone.
Introduction Advantages/ disadvantages Code examples Speed Summary Running on the AOD Analysis Platforms 1/11/2007 Andrew Mehta.
GAYA Analyzer SDD Presentation. GAYA Analyzer Introduction OMS40G256 is a hardware device used for detection of radioactive radiation for medical imaging.
Chapter 5: Process What Is Process?  Process Writing explains how to do something or describes how something is done. There Are Two Types of Process Paragraphs:
Principles of Computer Security: CompTIA Security + ® and Beyond, Third Edition © 2012 Principles of Computer Security: CompTIA Security+ ® and Beyond,
HBD HV Control System Development Manuel Proissl HBD Meeting 09/18/2007.
The Other Face Chapter 15. What documentation is required? ► Different levels of documentation are required for the casual user of a program, for the.
Making the System Operational Implementation & Deployment
© Copyright 2014 TONE SOFTWARE CORPORATION. Confidential and Proprietary. All rights reserved. ® Operator Training – Release Performance Dashboard.
1 Voicing Features Horacio Franco, Martin Graciarena Andreas Stolcke, Dimitra Vergyri, Jing Zheng STAR Lab. SRI International.
Oracle eBusiness Financials R12 Oracle Receivables Functional Overview TCS Oracle Practice.
IBM Software Group © 2008 IBM Corporation IBM Tivoli Provisioning Manager 7.1 Server Management/Task Management/Workflow.
Adaptive Software Development Process Framework. Version / 21 / 2001Page Project Initiation 2.0 Adaptive Cycle Planning 5.0 Final Q/A and.
Demand Planning Scenario Overview
Converting Third-Party Imaging Systems
Qifeng Zhu, Barry Chen, Nelson Morgan, Andreas Stolcke ICSI & SRI
District And Club database
5f. GSICS Wiki Overview and NOAA GSICS THREDDS Service Overview
Chapter 8 – Software Testing
Computer Structure Multi-Threading
BSA 376 Competitive Success/snaptutorial.com
BSA 376 Education for Service/snaptutorial.com
BSA 376 Teaching Effectively-- snaptutorial.com
Demand Planning Scenario Overview
Progress Report of Sphinx in Summer 2004 (July 1st to Aug 31st )
LTI Student Research Symposium 2004 Antoine Raux
Klopotek is transitioning to a Global Organization
Making the System Operational Implementation & Deployment
Guidance document for national employment flash estimates
Sphinx 3.X (X=4) Four-Layer Categorization Scheme of Fast GMM Computation Techniques in Large Vocabulary Continuous Speech Recognition Systems
Progress Report of Sphinx in Q (Sep 1st to Dec 30th)
Sphinx Recognizer Progress Q2 2004
Open Source Software Development Processes Version 2.5, 8 June 2002
Learning Long-Term Temporal Features
From STAB teleconference minutes
Setup QA Process Software Quality Assurance Telerik Software Academy
Presentation transcript:

CALO Decoder Progress Report for April/May Arthur (Decoder and ICSI Training) Jahanzeb (Decoder) Yitao (Decoder) Ziad (ICSI Training) Moss (ICSI Training) Carnegie Mellon University Apr 13, 2004

This Presentation (5 pages) Progress report for April/May In March Sphinx 3.4 is not ready Just start the script conversion for training In April/May Sphinx 3.4 is ready for release First-cut of AM and LM is done. (Thanks for Rita!)

Decoder Speed Accuracy Outlook in next 3 months (in 3.5) Compiler Optimization Doesn’t beat loop-unrolling However not using –D and using –ffast-math helps Phoneme-lookahead completes Accuracy Train a continuous HMM using all Communicator data. (S2 17% -> S3.4 14%) 64 mixtures will give us 12% ERR. Speed not-tuned. Outlook in next 3 months (in 3.5) WSJ: potential speed-up problem in task with > 5000 words. Speaker Adaptation: VTLN, MLLR, and techniques for fast enrollment Front-end transformation : LDA, HLDA, …… Model Combination experiment.

Decoder (Software) Release this week Not included (will be in 3.5) Mainly to replace buggy s3.3 Not included (will be in 3.5) Live mode APIs (Yitao , 80% completion) Outlook in next 3 months. Can learn from AHTK 1.3 Access of the models’ parameters? Server interface? Confidence measures?

ICSI Training Moss –LM training (done) Arthur/Ziad – 1 meeting training (done) Rita (Thanks!) – all meetings training (done) Current results (16 mixture, LM train for in-domain meeting) – 36.4% Different from the standard test set. Outlook in next 3 months Learn the magic from Rita, Wrap-up training script with perl. (Optional) Find a better test set. Start to improve the performance. With speaker adaptation technology. Class-based LM PLP

Infrastructure/Miscellaneous CVS is setup for MRCP ICSI conversion script Sphinx (will move back to Sourceforge.) Development is transitioning. Check-in training scripts? Scylla and Karybdis are running On a separate queue. Documentation for Sphinx. 3rd draft of outline completed. (9 chapters left.) …… () Outlook in next 3 months, Continue to maintain CVS. User education. Complete the 2nd draft of the documentation.

Outlook in next 3 months. Incorporate transform-based technology Speed-up for task > 5k words. Further improve ICSI training by all resources Transition development to CVS.