Lattice-Based Discriminative Training: Theory and Practice

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=83MMDJw8WsI


Duration: 48:04
805 views
10

Lattice-based discriminative training techniques such as MMI and MPE have been increasingly widely used in recent years.  I will review these model based discriminative training technique and also the newer feature-based techniques such as fMPE.  I will discuss some of the practical issues that are relevant to discriminative training, such as lattice generation, lattice depth and quality, probability scaling, I-smoothing, language models, alignment consistency, and various other issues for feature-based discriminative training, and will discuss more recent improvements such as frame-weighted MPE (MPFE), and give an overview of some recent unrelated work that I have been doing.  



Other Videos By Microsoft Research


2016-09-07Paying Attention to Interruption: A Human-Centered Approach to Intelligent Interruption Management
2016-09-07Studies of Programmers: How can they inform training and instruction?
2016-09-07Computers versus Common Sense [1/5]
2016-09-07Doing a Mao and the Xray Paradox: How Can Humanity Overcome Organization?           
2016-09-07THE CREATION: AN APPEAL TO SAVE LIVE ON EARTH
2016-09-07MobileASL: Making Cell Phones Accessible to the Deaf Community
2016-09-07Using Architecture and Code Optimization Techniques to Create Fast and Effective Data Compressors
2016-09-07A simple solution to the $k$-core problem
2016-09-07The Challenges of Development-through-Entrepreneurship: Research on Rural Computer Kiosks in India
2016-09-07The Elegant Solution: Toyota's Formula for Mastering Innovation
2016-09-07Lattice-Based Discriminative Training: Theory and Practice
2016-09-07Conference XP - Tutored Video Instruction With Conference XP and Classroom Presenter
2016-09-07Faster Decoding with Synchronous Grammars and n-gram Language Models
2016-09-07Locality and Phases: Dynamic Structures in Large-Scale Program Behavior
2016-09-07Inversion Transduction Grammar with Linguistic Constraints
2016-09-07How scheduling theory, scenarios, model checking and slicing can help in the verification of RTS
2016-09-07Innovention - the process of innovation and invention
2016-09-07Security and Privacy in Radio Frequency Identification
2016-09-07Conference XP - Automated Tracking of Student Behaviors
2016-09-07From Models to Systems: Applications of Model-based Design to Modern Large-Scale Systems
2016-09-07Splitting on Demand in Satisfiability Modulo Theories


Tags:
microsoft research