-
Notifications
You must be signed in to change notification settings - Fork 1
/
index.html
294 lines (227 loc) · 11.7 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
<!doctype html>
<html lang="en">
<head>
<meta name="google-site-verification" content="ftFOlJETX-2KNjaPh8W6s8lhigItRuu9fOmjHZZ0nY0" />
<!-- Required meta tags -->
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<!-- Bootstrap CSS -->
<link rel="stylesheet" href="https://stackpath.bootstrapcdn.com/bootstrap/4.1.3/css/bootstrap.min.css"
integrity="sha384-MCw98/SFnGE8fJT3GXwEOngsV7Zt27NXFoaoApmYm81iuXoPkFOJwJ8ERdknLPMO" crossorigin="anonymous">
<title>GraphSpeech</title>
</head>
<style type="text/css">
table {
width: 100%;
table-layout: fixed;
}
audio {
width: 100%;
}
thead>tr>th:first-child {
width: 96px;
}
@media (max-width: 767px) {
.big-screen {
display: none;
}
}
@media (min-width: 767px) {
.small-screen {
display: none;
}
}
</style>
<body>
<header class="header">
<div class="jumbotron bg-secondary text-center">
<div class="container">
<div class="row align-items-center">
<div class="col-md-12">
<h1><a class="text-light">GraphSpeech: Syntax-aware Graph Attention Network <br> For Neural Speech Synthesis</h1><br>
<font size=5><span style="color:#000000"> Authors: Rui Liu, Berrak Sisman, Haizhou Li </font> </a>
<p>
<div class="row">
</div>
</p>
</div>
</div>
</div>
</div>
</header>
<main>
<div class="container">
<div class="row" id="result">
<div class="col-md-12">
<h5>Abstract:</h5>
Attention-based end-to-end text-to-speech synthesis (TTS) is superior to conventional statistical methods in many ways. Transformer-based TTS is one of such successful implementations. While Transformer TTS models the speech frame sequence well with a self-attention mechanism, it does not associate input text with output utterances from a syntactic point of view at sentence level. We propose a novel neural TTS model, denoted as <i>GraphSpeech</i>, that is formulated under graph neural network framework. <i>GraphSpeech</i> encodes explicitly the syntactic relation of input lexical tokens in a sentence, and incorporates such information to derive syntactically motivated character embeddings for TTS attention mechanism. Experiments show that <i>GraphSpeech</i> consistently outperforms the Transformer TTS baseline in terms of spectrum and prosody rendering of utterances. <br>
<br>
<br>
<div style="text-align: center; width: 1100px; border: green solid 1px;">
<img align="center" src="model.png" width="600" height="520" />
<br>
<br>
Fig. 1: The proposed <i>GraphSpeech</i> that consists of Relation Encoder, Graph Encoder and Decoder. Its difference from Transformer TTS is highlighted in blue.
</div>
<div style="text-align: center; width: 1100px; border: green solid 1px;">
<img align="center" src="example.png" width="600" height="260" />
<br>
<br>
Fig. 2: An example of (a) syntax tree and (b) syntax graph of a sentence "I prefer the morning flight through Denver.". The blue words denote the dependency relation between two nodes. The black arrows denote directed one-way connections, and yellow arrows denote two-way connections. Self-loop connections are omitted due to space limitation.
</div>
<br><br><br>
<br>
<div class="big-screen">
<br>
<b> We use the state-of-the-art Transformer TTS <font color='blue'>[1]</font> as the baseline .</b>
<br>
<br>
<br /><br />
<h3>Speech Samples:</h3>
<br />
<br />
<!-- Naturalness Evaluation -->
<table class="table">
<tbody>
<tr>
<th></th>
<th style="text-align:center;"> Transformer TTS</th>
<th style="text-align:center;"> GraphSpeech </th>
</tr>
<tr>
<th colspan=4 align="center">[1] “We conduct objective evaluations to assess the performance of our proposed framework .”</th>
</tr>
<tr>
<th scope="row"> </th>
<td>
<audio controls>
<source src="samples/TT/1.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
<td>
<audio controls>
<source src="samples/GS/1.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
</tr>
<tr>
<th colspan=4 align="center">[2] “The main contributions of this paper are listed as follows .”</th>
</tr>
<tr>
<th scope="row"> </th>
<td>
<audio controls>
<source src="samples/TT/2.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
<td>
<audio controls>
<source src="samples/GS/2.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
</tr>
<tr>
<th colspan=4 align="center">[3] “Graphical structure plays an important role in natural language processing .”</th>
</tr>
<tr>
<th scope="row"> </th>
<td>
<audio controls>
<source src="samples/TT/3.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
<td>
<audio controls>
<source src="samples/GS/3.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
</tr>
<tr>
<th colspan=4 align="center">[4] “The blue words denote the dependency relation between two nodes .”</th>
</tr>
<tr>
<th scope="row"> </th>
<td>
<audio controls>
<source src="samples/TT/4.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
<td>
<audio controls>
<source src="samples/GS/4.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
</tr>
<tr>
<th colspan=4 align="center">[5] “We then encode them into vector representations .”</th>
</tr>
<tr>
<th scope="row"> </th>
<td>
<audio controls>
<source src="samples/TT/5.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
<td>
<audio controls>
<source src="samples/GS/5.wav" type="audio/wav" />
Your browser does not support the audio element.
</audio>
</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
<hr>
<div class="row" id="ref">
<div class="col">
<h5>References</h5>
[1] Naihan Li, Shujie Liu, Yanqing Liu, Sheng Zhao, and Ming Liu. Neural speech synthesis with transformer network. In AAAI 2019, pages 6706–6713.
<br>
</div>
</div>
<!--<hr>
<div class="row" id="contact">
<div class="col">
<h2>Contact</h2>
<div></div>
</div>
</div> -->
<div align="center" style="margin:auto;padding-top:10px">
<div style="width:20%">
<script type="text/javascript" id="clustrmaps" src="//clustrmaps.com/map_v2.js?d=ikRsfdaosDvlp5FPC8ESPOYxcatK16zJ5NDeYJ-r70I&cl=ffffff&w=a"></script>
</div>
</div>
<br><br>
</div>
</main>
<footer class="bg-secondary text-light mt-4 pt-3 pb-2 ">
<div class="container">
<p class="text-center">
</p>
</div>
</footer>
<!-- Optional JavaScript -->
<!-- jQuery first, then Popper.js, then Bootstrap JS -->
<script src="https://code.jquery.com/jquery-3.3.1.slim.min.js"
integrity="sha384-q8i/X+965DzO0rT7abK41JStQIAqVgRVzpbzo5smXKp4YfRvH+8abtTE1Pi6jizo"
crossorigin="anonymous"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.3/umd/popper.min.js"
integrity="sha384-ZMP7rVo3mIykV+2+9J3UJ46jBk0WLaUAdn689aCwoqbBJiSnjAK/l8WvCWPIPm49"
crossorigin="anonymous"></script>
<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.1.3/js/bootstrap.min.js"
integrity="sha384-ChfqqxuZUCnJSK3+MXmPNIyE6ZbWh2IMqE241rYiqJxyMiZ6OW/JmZQ5stwEULTy"
crossorigin="anonymous"></script>
</body>
</html>