500 INDEX
LRE (Language Resources and Evaluation)
program, 57
LSA (latent semantic analysis), 198
LSP (line spectral pairs), 229
LVCSR (large vocabulary continuous speech
recognition), 45, 81, 130, 266, 307
LVCSR (large vocabulary continuous speech
recognition)-based language
identification, 266–267
LVCSR (large vocabulary continuous speech
recognition)/HUB-5 technology
evaluations, 46
LVCSR (large vocabulary continuous speech
recognition)/Keyword Spotting, 269
M
machine translation, 195
machine translation evaluation, 383–389
automatic evaluation scheme, 387–389
overview, 383
rank evaluation method, 383–384
translation paired comparison method,
384–387
MAD (MT-Assisted Dialogs), 374–376
Mandarin Chinese, 51, 342–343
Maori language, 14
MAP (maximum a posteriori), 291
mapping, grapheme-to-phoneme, 29
Markov model, 251
maximum a posteriori (MAP), 291
maximum-likelihood (ML), 254
maximum-likelihood binary decision trees,
context clustering with, 256–259
maximum-likelihood linear regression
(MLLR), 291
MCC (Multi-Class Composite) bigram
model, 377
MDE (metadata extraction), 48
MDL-SSS model, 378
mel frequency cepstral coefficients (MFCC),
212, 229
metadata extraction (MDE), 48
MFCC (mel frequency cepstral coefficients),
212, 229
mixed-initiative dialog, 400
mixed-initiative interactions, 405
Mixer data collection, 52–55
Mixer Phase I, 53
Mixer robot operator, 53
ML (maximum-likelihood), 254
MLLR (maximum-likelihood linear
regression), 291
ML-mix combination method, 107
ML-mix method, 106
ML-sep combination method, 107
ML-sep method, 106
ML-tag combination method, 107
model estimation for new domains and speaking
styles, 174–177
language model adaptation, 176–177
overview, 174–176
models for new languages and multilingual
applications, 173–174
Modern Standard Arabic (MSA), 38, 191
Moldovan, 6–7
monolingual recognizer, 77
monolingual SDS implemented in VXML, 439
monophonemes, 93
morphemes, 16
Morphological Analyzer, 336
morphologically rich languages modeling,
180–188
factored language models, 185–186
free word order difficulties, 187–188
hypothesis driven lexical adaptation,
186–187
morphs derived from linear decomposition of
words as units of language modeling,
183–185
overview, 180–183
problems of very large vocabulary, 183
morphological richness, 178
morphology, 16–19, 87–89
morphs, 16, 183–185
MSA (Modern Standard Arabic), 38, 191
MT-Assisted Dialogs (MAD), 374–376
Multi-Class Composite (MCC) bigram
model, 377
multilingual acoustic modeling, 71–122
acoustic model combination, 102–118
approaches based on alternative sound
units, 112–118
language adaptive acoustic modeling,
110–112
language independent acoustic modeling,
102–110
overview, 102

Get Multilingual Speech Processing now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.