“Improving Pronunciation Dictionary Coverage of Names by Modelling Spelling Variation” - Justin Fackrell and Wojciech Skut Presented by Han.

Slides:



Advertisements
Similar presentations
Recommender Systems & Collaborative Filtering
Advertisements

Spelling Correction for Search Engine Queries Bruno Martins, Mario J. Silva In Proceedings of EsTAL-04, España for Natural Language Processing Presenter:
Random Forest Predrag Radenković 3237/10
Speech Recognition Part 3 Back end processing. Speech recognition simplified block diagram Speech Capture Speech Capture Feature Extraction Feature Extraction.
Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data John Lafferty Andrew McCallum Fernando Pereira.
Development of Automatic Speech Recognition and Synthesis Technologies to Support Chinese Learners of English: The CUHK Experience Helen Meng, Wai-Kit.
Using Contiguous Bi-Clustering for data driven temporal analysis of fMRI based functional connectivity.
Commands and predicates LISP functions are divided into 2 classes. Predicates are functions that return boolean values i.e. t or nil. The rest are commands.
Feature Selection for Regression Problems
Semantic text features from small world graphs Jure Leskovec, IJS + CMU John Shawe-Taylor, Southampton.
Predicting the Semantic Orientation of Adjective Vasileios Hatzivassiloglou and Kathleen R. McKeown Presented By Yash Satsangi.
Transformation-based error- driven learning (TBL) LING 572 Fei Xia 1/19/06.
Predicting the Semantic Orientation of Adjectives
Gobalisation Week 8 Text processes part 2 Spelling dictionaries Noisy channel model Candidate strings Prior probability and likelihood Lab session: practising.
The Implicit Mapping into Feature Space. In order to learn non-linear relations with a linear machine, we need to select a set of non- linear features.
Building High Quality Databases for Minority Languages such as Galician F. Campillo, D. Braga, A.B. Mourín, Carmen García-Mateo, P. Silva, M. Sales Dias,
WALT: To recognise and extend number sequences.
L. Padmasree Vamshi Ambati J. Anand Chandulal J. Anand Chandulal M. Sreenivasa Rao M. Sreenivasa Rao Signature Based Duplicate Detection in Digital Libraries.
Where Innovation Is Tradition SYST699 – Spec Innovations Innoslate™ System Engineering Management Software Tool Test & Analysis.
Face Model Fitting with Generic, Group-specific, and Person- specific Objective Functions Chair for Image Understanding and Knowledge-based Systems Institute.
Protein Sequence Alignment and Database Searching.
Week 9: resources for globalisation Finish spell checkers Machine Translation (MT) The ‘decoding’ paradigm Ambiguity Translation models Interlingua and.
Classification. An Example (from Pattern Classification by Duda & Hart & Stork – Second Edition, 2001)
Evaluating Statistically Generated Phrases University of Melbourne Department of Computer Science and Software Engineering Raymond Wan and Alistair Moffat.
Globalisation and machine translation Machine Translation (MT) The ‘decoding’ paradigm Ambiguity Translation models Interlingua and First Order Predicate.
1 Cross-Lingual Query Suggestion Using Query Logs of Different Languages SIGIR 07.
Learning Phonetic Similarity for Matching Named Entity Translation and Mining New Translations Wai Lam, Ruizhang Huang, Pik-Shan Cheung ACM SIGIR 2004.
Copyright 2007, Toshiba Corporation. How (not) to Select Your Voice Corpus: Random Selection vs. Phonologically Balanced Tanya Lambert, Norbert Braunschweiler,
SLTU 2014 – 4th Workshop on Spoken Language Technologies for Under-resourced Languages St. Petersburg, Russia KIT – University of the State.
Avoiding Segmentation in Multi-digit Numeral String Recognition by Combining Single and Two-digit Classifiers Trained without Negative Examples Dan Ciresan.
12-6 Rational Expressions with Like Denominators Objective: Students will be able to add and subtract rational expressions with like denominators.
Mining Reference Tables for Automatic Text Segmentation Eugene Agichtein Columbia University Venkatesh Ganti Microsoft Research.
Lesson 8-6B Use Cube Roots and Fractional Exponents After today’s lesson, you should be able to evaluate cube roots and simplify expressions with fractional.
Copyright 2007, Paradigm Publishing Inc. ACCESS 2007 Chapter 3 BACKNEXTEND 3-1 LINKS TO OBJECTIVES Modify a Table – Add, Delete, Move Fields Modify a Table.
L ETTER TO P HONEME A LIGNMENT Reihaneh Rabbany Shahin Jabbari.
Latent Semantic Mapping (LSA) 資工四 阮鶴鳴 資工四 李運寰 “A multi-span language modeling framework for large vocabulary speech recognition,” J.R. Bellegarda -, 1998.
We are learning to write expressions using variables. (8-1)
Pop Quiz – Self assessment What is the difference between… – decoding and encoding – phonemes and graphemes – sight words and decoded words – phonological.
Improving Named Entity Translation Combining Phonetic and Semantic Similarities Fei Huang, Stephan Vogel, Alex Waibel Language Technologies Institute School.
Learning Phonetic Similarity for Matching Named Entity Translations and Mining New Translations Wai Lam Ruizhang Huang Pik-Shan Cheung Department of Systems.
Letter to Phoneme Alignment Using Graphical Models N. Bolandzadeh, R. Rabbany Dept of Computing Science University of Alberta 1 1.
The UNIVERSITY of NORTH CAROLINA at CHAPEL HILL Classification COMP Seminar BCB 713 Module Spring 2011.
John Lafferty Andrew McCallum Fernando Pereira
Decision Tree Learning Presented by Ping Zhang Nov. 26th, 2007.
Generating Query Substitutions Alicia Wood. What is the problem to be solved?
Objectives: Students will be able to… Use properties of rational exponents to evaluate and simplify expressions Use properties of rational exponents to.
© Hamilton Trust Keeping Up Term 2 Week 2 Day 1 Objectives: Know by heart pairs with a total of every number up to 20 Add three single digit numbers using.
Math 7  Bellwork #2  Please have the following on your desk: Name Tag Math book HW Red Pen Pencil.
Machine Learning in Practice Lecture 2 Carolyn Penstein Rosé Language Technologies Institute/ Human-Computer Interaction Institute.
2 2.5 © 2016 Pearson Education, Ltd. Matrix Algebra MATRIX FACTORIZATIONS.
How Spelling Supports Reading Based on the article “Why Spelling Supports Reading And Why It Is More Regular and Predictable Than You May Think” By Louisa.
Enabling technology Enabling what and for whom?. This workshop… We will use ICT-tools for specific purposes. We want to transform the way we do things.
By Shaunte Morris. Galena Park Middle School football team made one 6-point touchdown and four 3-point field goals in their last game. Shanique and Justin.
Objectives: 1)Students will be able to find the inverse of a function or relation. 2)Students will be able to determine whether two functions or relations.
Building Learner Profiles for EFL Reading Intervention through the Use of the ABLE (Assessing Basic Literacy in English) Kit Dr. Susie Russak, Beit Berl.
Cognitive Play: Smart Mouth
Speaker : chia hua Authors : Long Qin, Ming Sun, Alexander Rudnicky
Do-Gil Lee1*, Ilhwan Kim1 and Seok Kee Lee2
Word Pronunciation Julia Hirschberg CS /18/2018.
Pronouncing Words in TTS Systems
Lesson Four: Building Custom Patient Lists
Sequences COURSE 3 LESSON 12-1
ورود اطلاعات بصورت غيربرخط
Rohit Kumar *, Amit Kataria, Sanjeev Sofat
Word Pronunciation Julia Hirschberg 4/14/2019.
BEFORE YOU BEGIN Read the Word Sort Directions
Trees in java.util A set is an object that stores unique elements
Relations/Sequences Objective: Students will learn how to identify if a relation is a function. They will also be able to create a variable expression.
English phonetic symbols (Consonant Symbols)
Final Exam Grading This is a skills class, you will be graded differently than your other classes.
Presentation transcript:

“Improving Pronunciation Dictionary Coverage of Names by Modelling Spelling Variation” - Justin Fackrell and Wojciech Skut Presented by Han

The Problem: The pronunciation of out-of-vocabulary (OOV) words is a major problem in TTS. Many OOV words are names. For English names, the orthography for names is highly irregular. Current methods of approaching this problem has low accuracy. –Using hand-written or automatically learned rules to replace a sequence of graphemes by a sequence of phonemes.

The Challenge

Their Method Scope: English surnames, forenames, street names and place names. Based on: the observation that some of the words in the above categories have same pronunciation, but slightly different spelling. Approach: learn from existing data (data- driven) of the rules of these variations, so that next time we see an OOV word, we will try to apply these rules and see if we can transform that word into an IOV word.

Different Orthographical Expressions for the Same Pronunciation

Hypothesis Given a name that’s not in the dictionary, there’s about 10% chance that it DOES have a valid pronunciation in the dictionary. We have to somehow map it to a valid in-dictionary word.

A Hierarchical Approach Dictionary Filter 1 Filter 2 etc.

Two Ways of Using This Method and Their Results Online –Results suggested pronunciations are good in 80% of cases. Offline –For surnames, a model trained on a 23,000- entry dictionary was able to add 5,000 new entries, increasing the coverage by about 1%.

The Algorithm (Part I) Training 1) reverse dictionary (pron -> ortho) 2) delete one-to-one mappings 3) Each pair of spellings that share a common pronunciation generates a set of rewrite rules, r i where i = 0 to n, in the form of “A -> B / L _ R”

The Algorithm (Part I) Training

Each rule, r i, is then evaluated on the rest of the dictionary to see how useful it is. –MISS –OOV –DIFF –GOOD And gets four scores: n i MISS, n i OOV, n i DIFF, and n i GOOD From each set of rules generated by a pair, only one rule is chosen: shortest and n i DIFF =0.

The Algorithm (Part I) Predication Sort all rules by score. When given an OOV word, use the rule with the highest score that can map it into an IOV word.

Some Examples of Resulted Rewrite Rules

Some Results

Accuracy Test Results