Your browser does not support the video tag. Please use IE9+ or Google Chrome.

01:21
1. Slide 109:47
2. Possible Candidatesphrases, words, syllables, phonemes.....Phonemethe minimum units of speech sound in a language which can serve to distinguish one word from the othere.g. b at / p at , b a d / b e dphone : a phoneme’s acoustic realization the same phoneme may have many different realizations e.g. sa t / me t erCoarticulation and Context Dependencycontext: right/left neighboring unitscoarticulation: sound production changed because of the neighboring unitsright-context-dependent (RCD)/left-context-dependent (LCD)/ bothintraword/interword context dependencyFor Mandarin Chinesecharacter/syllable mapping relationsyllable: Initial (聲母) / Final (韻母) / tone (聲調)03:48
3. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:03
4. Possible Candidatesphrases, words, syllables, phonemes.....Phonemethe minimum units of speech sound in a language which can serve to distinguish one word from the othere.g. b at / p at , b a d / b e dphone : a phoneme’s acoustic realization the same phoneme may have many different realizations e.g. sa t / me t erCoarticulation and Context Dependencycontext: right/left neighboring unitscoarticulation: sound production changed because of the neighboring unitsright-context-dependent (RCD)/left-context-dependent (LCD)/ bothintraword/interword context dependencyFor Mandarin Chinesecharacter/syllable mapping relationsyllable: Initial (聲母) / Final (韻母) / tone (聲調)02:52
5. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:12
6. Possible Candidatesphrases, words, syllables, phonemes.....Phonemethe minimum units of speech sound in a language which can serve to distinguish one word from the othere.g. b at / p at , b a d / b e dphone : a phoneme’s acoustic realization the same phoneme may have many different realizations e.g. sa t / me t erCoarticulation and Context Dependencycontext: right/left neighboring unitscoarticulation: sound production changed because of the neighboring unitsright-context-dependent (RCD)/left-context-dependent (LCD)/ bothintraword/interword context dependencyFor Mandarin Chinesecharacter/syllable mapping relationsyllable: Initial (聲母) / Final (韻母) / tone (聲調)00:22
7. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:23
8. Possible Candidatesphrases, words, syllables, phonemes.....Phonemethe minimum units of speech sound in a language which can serve to distinguish one word from the othere.g. b at / p at , b a d / b e dphone : a phoneme’s acoustic realization the same phoneme may have many different realizations e.g. sa t / me t erCoarticulation and Context Dependencycontext: right/left neighboring unitscoarticulation: sound production changed because of the neighboring unitsright-context-dependent (RCD)/left-context-dependent (LCD)/ bothintraword/interword context dependencyFor Mandarin Chinesecharacter/syllable mapping relationsyllable: Initial (聲母) / Final (韻母) / tone (聲調)01:23
9. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques01:16
10. Sharing of Parameters and Training Data for Triphones00:09
11. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:15
12. Sharing of Parameters and Training Data for Triphones00:20
13. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:14
14. Sharing of Parameters and Training Data for Triphones00:21
15. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:09
16. Sharing of Parameters and Training Data for Triphones00:04
17. Some Fundamentals in Information Theory00:05
18. Sharing of Parameters and Training Data for Triphones00:40
19. Primary Considerationsaccuracy: accurately representing the acoustic realizationstrainability: feasible to obtain enough data to estimate the model parametersgeneralizability: any new word can be derived from a predefined unit inventoryExampleswords: accurate if enough data available, trainable for small vocabulary, NOT generalizablephoneme : trainable, generalizable difficult to be accurate due to context dependencysyllable: 50 in Japanese, 1300 in Mandarin Chinese, over 30000 in English Triphonea phoneme model taking into consideration both left and right neighboring phonemes(60)3→ 216,000very good generalizability, balance between accuracy/ trainability by parameter-sharing techniques00:01
20. Sharing of Parameters and Training Data for Triphones00:06
21. Some Fundamentals in Information Theory00:01
22. Fundamentals in Information Theory00:02
23. Some Fundamentals in Information Theory00:03
24. Fundamentals in Information Theory00:01
25. Fundamentals in Information Theory00:02
26. Fundamentals in Information Theory00:02
27. Some Fundamentals in Information Theory00:06
28. Classification and Regression Trees (CART)00:02
29. Goal00:01
30. Splitting Criteria for the Decision Tree00:01
31. Training Triphone Models with Decision Trees00:07
32. Training Tri-phone Models with Decision Trees00:55
33. Some Fundamentals in Information Theory00:04
34. Fundamentals in Information Theory00:05
35. Some Fundamentals in Information Theory00:06
36. Fundamentals in Information Theory00:04
37. Some Fundamentals in Information Theory00:11
38. Fundamentals in Information Theory00:07
39. Some Fundamentals in Information Theory00:17
40. Fundamentals in Information Theory00:15
41. Some Fundamentals in Information Theory00:06
42. Fundamentals in Information Theory00:22
43. Some Fundamentals in Information Theory00:03
44. Fundamentals in Information Theory00:02
45. Some Fundamentals in Information Theory00:01
46. Fundamentals in Information Theory00:29
47. Some Fundamentals in Information Theory00:25
48. Fundamentals in Information Theory00:20
49. Some Fundamentals in Information Theory00:52
50. Fundamentals in Information Theory00:20
51. Some Fundamentals in Information Theory00:50
52. Fundamentals in Information Theory01:12
53. Some Fundamentals in Information Theory00:19
54. Fundamentals in Information Theory01:44
55. Some Fundamentals in Information Theory00:53
56. Some Fundamentals in Information Theory00:04
57. Fundamentals in Information Theory00:12
58. Some Fundamentals in Information Theory00:33
59. Fundamentals in Information Theory00:15
60. Some Fundamentals in Information Theory00:11
61. Fundamentals in Information Theory00:13
62. Some Fundamentals in Information Theory00:07
63. Fundamentals in Information Theory00:58
64. Some Fundamentals in Information Theory00:02
65. Fundamentals in Information Theory00:12
66. Some Fundamentals in Information Theory00:02
67. Fundamentals in Information Theory00:09
68. Some Fundamentals in Information Theory00:02
69. Fundamentals in Information Theory00:49
70. Some Fundamentals in Information Theory00:02
71. Fundamentals in Information Theory00:04
72. Some Fundamentals in Information Theory00:01
73. Fundamentals in Information Theory02:15
74. Some Fundamentals in Information Theory00:03
75. Fundamentals in Information Theory00:10
76. Some Fundamentals in Information Theory00:53
77. Fundamentals in Information Theory00:06
78. Some Fundamentals in Information Theory01:29
79. Fundamentals in Information Theory01:14
80. Fundamentals in Information Theory00:09
81. Fundamentals in Information Theory01:58
82. Fundamentals in Information Theory00:32
83. Fundamentals in Information Theory00:05
84. Fundamentals in Information Theory00:10
85. Fundamentals in Information Theory00:03
86. Fundamentals in Information Theory00:26
87. Fundamentals in Information Theory00:05
88. Fundamentals in Information Theory00:03
89. Fundamentals in Information Theory00:01
90. Fundamentals in Information Theory00:13
91. Fundamentals in Information Theory00:02
92. Some Fundamentals in Information Theory00:01
93. Fundamentals in Information Theory00:02
94. Fundamentals in Information Theory00:12
95. Fundamentals in Information Theory00:01
96. Fundamentals in Information Theory01:40
97. Fundamentals in Information Theory00:03
98. Fundamentals in Information Theory00:02
99. Fundamentals in Information Theory00:01
100. Fundamentals in Information Theory01:46
101. Fundamentals in Information Theory00:01
102. Fundamentals in Information Theory00:01
103. Fundamentals in Information Theory00:01
104. Some Fundamentals in Information Theory00:01
105. Fundamentals in Information Theory00:31
106. Some Fundamentals in Information Theory00:01
107. Fundamentals in Information Theory00:01
108. Some Fundamentals in Information Theory00:07
109. Fundamentals in Information Theory00:04
110. Fundamentals in Information Theory00:07
111. Fundamentals in Information Theory00:11
112. Some Fundamentals in Information Theory00:01
113. Fundamentals in Information Theory00:01
114. Fundamentals in Information Theory00:01
115. Fundamentals in Information Theory00:01
116. Some Fundamentals in Information Theory00:01
117. Fundamentals in Information Theory00:10
118. Some Fundamentals in Information Theory00:01
119. Fundamentals in Information Theory00:01
120. Some Fundamentals in Information Theory00:01
121. Fundamentals in Information Theory00:01
122. Fundamentals in Information Theory00:01
123. Fundamentals in Information Theory01:16
124. Some Fundamentals in Information Theory00:07
125. Fundamentals in Information Theory01:15
126. Some Fundamentals in Information Theory00:32
127. Fundamentals in Information Theory00:01
128. Fundamentals in Information Theory00:01
129. Fundamentals in Information Theory00:02
130. Some Fundamentals in Information Theory04:45
131. Classification and Regression Trees (CART)00:03
132. Goal00:09
133. Classification and Regression Trees (CART)00:24
134. Goal00:01
135. Classification and Regression Trees (CART)00:01
136. Some Fundamentals in Information Theory00:12
137. Fundamentals in Information Theory00:01
138. Some Fundamentals in Information Theory00:01
139. Classification and Regression Trees (CART)01:35
140. Goal00:02
141. Classification and Regression Trees (CART)00:59
142. Goal00:01
143. Classification and Regression Trees (CART)00:01
144. Some Fundamentals in Information Theory00:10
145. Fundamentals in Information Theory00:01
146. Some Fundamentals in Information Theory00:01
147. Classification and Regression Trees (CART)00:33
148. Goal00:04
149. Splitting Criteria for the Decision Tree00:21
150. Goal00:15
151. Splitting Criteria for the Decision Tree00:01
152. Goal00:10
153. Classification and Regression Trees (CART)00:39
154. Goal