20
An Efficient Distance Measure for Comparing Phoneme Sequences Baruchi Har-Lev Speech Recognition Day 8/6/2010

An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Embed Size (px)

Citation preview

Page 1: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

An Efficient Distance Measure for

Comparing Phoneme Sequences

Baruchi Har-Lev Speech Recognition Day – 8/6/2010

Page 2: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Multi-Stage LVCSR

2

Vocabulary

Reduction

Phoneme

Decoder

Lexicon

CSR

Input

Speech

Textual

Output

Probable

Word

Selection

Recognized

Phoneme

Sequence

Reduced

Vocabulary

Distance

Measure

Hypothesis

Creation

Search Space Creation

Lexicon

Page 3: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Distance measure

O(n2)

Overview The Problem

3

Word1

.

.

Word100k

Ph1 Ph2 . Phk . PhM

Example

Lexicon size = 100k

Phoneme sequence length = 250

Grid size = 25M

Avg. hypothesis length = 7

Number of Operations > 1,000,000,000

Recognized Phoneme

Sequence Lexicon

Wordi

Page 4: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Reduce computational complexity of the

distance measure while maintaining

minimum performances loss

Overview

The Goal

4

Page 5: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Test Data Base

5

VoiceMail Macrophone Parameter

2485 3140 # of Utterances

20+ sec 3.5 sec Utterance Length

70-80 8-9 # of Words in Utterance

250 40 # of Phonemes in

Utterance

Page 6: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Insertion, Deletion, Substitution

Equal weight probabilities

O(NM) ≈ O(n2)

Levenshtein Distance

6

D

ji PhPhSub ,

1jPh jPh

1iPh

iPh jPhIns

iPhDel

Ref.

Phoneme

String

Tested Phoneme

String

InsDelSubMinPhPhD ji ,,,

Page 7: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Levenshtein Distance in the

Exhaustive Process

7

.

X1

XN

.

0

Y1 . . YM

Lexicon

Word

Hypothesis

from Sequence

W1

.

.

.

W100k

Y T N . . .

Full Search Grid

2

:,

nOSizeGrid

HypWDofNumber ji

Page 8: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Results

8

Coverage [%] Creation Time

[sec] Distance Method

85% 8 Levenshtein Distance

Reduced

Vocabulary Size

Lexicon

Size

Mean Sequence

Size [Phones] DB

50k 100k 41 Macrophone

Page 9: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Choose a canonic word - Word0

Compute distances (Levenshtein)

A-priori distance computation

Real time computation

Sum of results

A-Priori Distance Differences

9

HypWordDWordWiDHypWiD ,,, 00

A-priori One Time in RT

Page 10: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

A-Priori Distance Differences Full Search

10

O(Grid Size)

priori-A RT

W1

.

.

.

W100k

1W

0,WordWiD

kW100

Hyp1

.

.

.

HypN

1Hyp

0,WordHypD i

NHyp

+

O(#Hyp) O(1)

Page 11: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Results

11

Coverage [%] Creation Time

[sec] Distance Method

85% 8 Levenshtein Distance

53% 0.32 A-priori Differences

Reduced

Vocabulary Size

Lexicon

Size

Mean Sequence

Size [Phones] DB

50k 100k 41 Macrophone

96% Time Reduction, 32% Coverage Loss

Page 12: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Advantages:

Reduces number of

hypotheses

“Breaks” the distance

measure

Disadvantages:

Complex interface

Tree Distance

12

t

L4

“about”

v

L4

“above”

m

L4

“charm”

ey1

L2

{}

ax

L1

“a”

Ch

L1

{}

aw1

L3

{}

ah1

L3

{}

aa1

L2

{}

0

r

L3

{}

n

L3

“chain”

b

L2

{}

Page 13: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Results

13

Coverage [%] Creation Time

[sec] Distance Method

85% 8 Levenshtein Distance

53% 0.32 A-priori Differences

85% 8 Tree Distance

Reduced

Vocabulary Size

Lexicon

Size

Mean Sequence

Size [Phones] DB

50k 100k 41 Macrophone

No Improvement

Page 14: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Stringology – Computer science

O(n1.4N1.2)

Good compression rate

Avg. length (Lexicon Word) ≈ 7 Phones

Bad compression rate

Biology

Blast

Computer network

A Glimpse at Different Disciplines

14

Page 15: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Length(Hyp.) = Length(Word)

Basic comparisons

Uses substitutions only

Does not use:

Diagonal Distance

15

.

X1

XN

.

0

Y1 . . YM

Lexicon

Word

Hypothesis from

Sequence

Insertion Deletion

Page 16: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Results

16

Coverage [%] Creation Time

[sec] Distance Method

85% 8 Levenshtein Distance

53% 0.32 A-priori Differences

85% 8 Tree Distance

83% 0.9 Diagonal Distance

Reduced

Vocabulary Size

Lexicon

Size

Mean Sequence

Size [Phones] DB

50k 100k 41 Macrophone

90% Time Reduction, 2% Coverage Loss

Page 17: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Additional data - Phoneme engine characteristics

Confusion matrix

Posteriori probabilities

Larger distinction between

phones

Disadvantage

Sensitivity to phoneme

engine characteristics Del

Ins

.

Ph1

0

Ph1 . . PhN

PhN

.

i

j

Weighted Distance

17

Page 18: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Results

18

Coverage [%] Creation Time

[sec] Distance Method

85% 8 Levenshtein Distance

53% 0.32 A-priori Differences

85% 8 Tree Distance

83% 0.9 Diagonal Distance

90% 0.9 Diagonal Weighted Distance

Reduced

Vocabulary Size

Lexicon

Size

Mean Sequence

Size [Phones] DB

50k 100k 41 Macrophone

90% Time Reduction, 5% Coverage Improvement

Page 19: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

Vocabulary reduction

Problems and Goals

Levenshtein distance method

Very complex in exhaustive search

Improving the distance method

Best method – Weighted diagonal distance method

Next step

Summary

19

Page 20: An Efficient Distance Measure for Comparing Phoneme Sequencesevents.eventact.com/afeka/afeka/An Efficient Distance Measure for... · An Efficient Distance Measure for Comparing Phoneme

20