forked from tlk1997/demo
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathintroduction.html
346 lines (284 loc) · 18 KB
/
introduction.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
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=0, minimal-ui">
<title>DeepKE</title>
<meta content="DeepKE" name="description" />
<meta content="ZJU-KLab" name="author" />
<meta http-equiv="X-UA-Compatible" content="IE=edge" />
<link rel="shortcut icon" href="assets/images/zj.png">
<!-- jvectormap -->
<link href="assets/plugins/jvectormap/jquery-jvectormap-2.0.2.css" rel="stylesheet">
<link href="assets/plugins/fullcalendar/vanillaCalendar.css" rel="stylesheet" type="text/css" />
<link href="assets/plugins/morris/morris.css" rel="stylesheet">
<link href="assets/css/bootstrap.min.css" rel="stylesheet" type="text/css">
<link href="assets/css/icons.css" rel="stylesheet" type="text/css">
<link href="assets/css/style.css" rel="stylesheet" type="text/css">
<link href="assets/plugins/alertify/css/alertify.css" rel="stylesheet" type="text/css">
</head>
<body class="fixed-left">
<!-- Loader -->
<div id="preloader"><div id="status"><div class="spinner"></div></div></div>
<!-- Begin page -->
<div id="wrapper">
<!-- ========== Left Sidebar Start ========== -->
<div class="left side-menu" style="width: 320px;">
<button type="button" class="button-menu-mobile button-menu-mobile-topbar open-left waves-effect">
<i class="ion-close"></i>
</button>
<!-- LOGO -->
<div class="topbar-left">
<div class="text-center">
<a href="https://github.com/zjunlp/deepke" class="logo">
<img src="assets/images/logo_openkg.png" alt="" class="logo-large" style="height: 70px;">
</a>
</div>
</div>
<div class="sidebar-inner niceScrollleft">
<div id="sidebar-menu">
<ul>
<li class="menu-title">全监督</li>
<li>
<a href="ner_st.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 常规实体抽取 </span>
</a>
</li>
<li>
<a href="ae_st.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 常规属性抽取</span>
</a>
</li>
<li>
<a href="re_st.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 常规关系抽取</span>
</a>
</li>
</ul>
<ul>
<li class="menu-title">低资源</li>
<li>
<a href="a1.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 少样本实体抽取 </span>
</a>
</li>
<li>
<a href="a2.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 少样本关系抽取 </span>
</a>
</li>
</ul>
<ul>
<li class="menu-title">篇章级</li>
<li>
<details open style="text-indent: 26px;cursor: pointer;-webkit-tap-highlight-color: transparent;">
<summary style="padding: 15px 20px;letter-spacing: .035em;cursor: default;font-size: 15px;color: rgb(71, 71, 71);">文档级关系抽取</summary>
<li>
<a href="welcome.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 展示样例</span>
</a>
</li>
<br/>
<li>
<a href="input.html" class="waves-effect">
<i class="mdi mdi-code-not-equal"></i>
<span> 输入样例</span>
</a>
</li>
<br/>
</details>
</li>
</ul>
</div>
<div class="clearfix"></div>
</div> <!-- end sidebarinner -->
</div>
<div class="content-page" style="margin-left: 320px">
<div class="content">
<!-- Top Bar Start -->
<div class="topbar">
<nav class="navbar-custom" style="padding-right: 13px">
<ul class="list-inline float-right mb-0">
<li class="list-inline-item dropdown notification-list" style="padding: 10px;">
<a href="https://azft.alibaba.com/lab/?id=7" role="button >
aria-haspopup="false" aria-expanded="false" >
<img src="assets/images/zjus.jpeg" alt="user" class="rounded-circle" style="height: 50px;width: auto;">
</a>
</li>
</ul>
<ul class="list-inline menu-left mb-0">
<li class="float-left">
<button class="button-menu-mobile open-left waves-light waves-effect">
<i class="mdi mdi-menu"></i>
</button>
</li>
</ul>
<div class="clearfix">
<ul>
<li><a href="index.html">DeepKE</a> </li>
<li><a href="introduction.html">介绍</a> </li>
<li><a href="https://github.com/zjunlp/deepke">下载</a> </li>
<li><a href="connection.html">联系我们</a> </li>
<li><a href="http://openkg.cn/">OpenKG</a></li>
</ul>
</div>
</nav>
</div>
<!-- Top Bar End -->
<div class="page-content-wrapper ">
<div class="container-fluid">
<div class="row" id="div_head_websites">
<div class="col-sm-12">
<div class="page-title-box">
<div class="btn-group float-right">
<ol class="breadcrumb hide-phone p-0 m-0">
</ol>
</div>
<h4 class="page-title"></h4>
</div>
</div>
<div class="clearfix">
</div>
</div>
<div class="row" id="div_chosen">
<div class="col-sm-12">
<div class="card">
<div class="card-body active">
<h5 class="header-title mt-0 mb-3">1.DeepKE介绍</h5>
<p class="mb-0" id="paragraph_1">DeepKE 是一个支持低资源、长篇章的知识抽取工具，可以基于pytorch实现命名实体识别、关系抽取和属性抽取功能。</p>
<br/>
<h5 class="header-title mt-0 mb-3">2.DeepKE模型架构</h5>
<p class="mb-0">Deepke的架构图如下所示</p>
<img src="assets/images/model1.png" alt="" style="height: auto; width: auto\9; width:100%;">
<p class="mb-0">DeepKE包括了三个模块，可以进行关系抽取、实体命名识别以及属性抽取的任务，在各个模块下包括各自的子模块。其中关系抽取模块就有常规模块、文档级抽取模块以及低资源少样本模块。在每一个子模块中，分为分词、预处理等的一个工具集合，以及编码和训练预测部分。</p>
<br/>
<h5 class="header-title mt-0 mb-3">3.DeepKE快速上手</h5>
<h6>(1)命名实体识别NER</h6>
<p class="mb-0">数据为txt文件，样式范例为：</p>
<img src="assets/images/tu1.png" alt="" style="height: auto; width: auto\9; width:100%;">
<br/>
<p class="mb-0">具体流程请进入详细的README中：</p>
<br/>
<p class="mb-0"><a href="https://github.com/zjunlp/deepke/blob/test_new_deepke/example/ner/standard">STANDARD</a> ： 常规模块为预训练模型，可进入其目录，修改数据集以及conf文件夹下的目录，python run.py即可训练，python predict.py即可预测。</p>
<p class="mb-0"><a href="https://github.com/zjunlp/DeepKE/tree/test_new_deepke/example/ner/few-shot">FEW-SHOT </a> ： 少样本模块使用了LightNER模型，可进入其目录，模型加载和保存位置以及配置可以在shell脚本中修改python run.py训练conll2003,python run.py +train=few_shot直接进行few-shot训练,若要加载模型，修改few_shot.yaml中的load_path,python predict.py即可预测。</p>
<h6>(2)关系抽取RE</h6>
<p class="mb-0">关系抽取是从非结构化的文本中抽取出实体之间的关系，以下为几个样式范例，数据为csv文件：</p>
<img src="assets/images/tu2.png" alt="" style="height: auto; width: auto\9; width:100%;">
<p class="mb-0">具体流程请进入详细的README中，RE包括了以下三个子功能:</p>
<br/>
<p class="mb-0"><a href="https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/standard">STANDARD</a> ： 常规模块为常用的深度学习模型，有CNN、RNN、Capsule、GCN、Transforemer以及预训练模型，可进入其目录，修改数据集以及conf文件夹下的目录，python run.py即可训练，python predict.py即可预测。</p>
<p class="mb-0"><a href="https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/few-shot">FEW-SHOT </a> : 进入其目录，修改数据集以及conf文件夹下的目录，python run.py即可训练，如需从上次训练的模型开始训练：设置.yaml中的train_from_saved_model为上次保存模型的路径，每次训练的日志保存路径默认保存在根目录，可以通过.yaml中的log_dir来配置python predict.py即可预测。</p>
<p class="mb-0"><a href="https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/document">DOCUMENT </a> : train_distant.json由于文件太大，请自行从Google Drive上下载到data/目录下.进入其目录，修改数据集以及conf文件夹下的目录，python run.py即可训练，如需从上次训练的模型开始训练：设置.yaml中的train_from_saved_model为上次保存模型的路径，每次训练的日志保存路径默认保存在根目录，可以通过.yaml中的log_dir来配置python predict.py即可预测。</p>
<h6>(3)属性抽取AE</h6>
<p class="mb-0">数据为csv文件，样式范例为：</p>
<img src="assets/images/tu3.png" alt="" style="height: auto; width: auto\9; width:100%;">
<p class="mb-0">具体流程请进入详细的README中:</p>
<br/>
<p class="mb-0"><a href="https://github.com/zjunlp/deepke/blob/test_new_deepke/example/ae/standard">STANDARD</a> ： 常规模块为常用的深度学习模型，有CNN、RNN、Capsule、GCN、Transforemer以及预训练模型，可进入其目录，修改数据集以及conf文件夹下的目录，python run.py即可训练，python predict.py即可预测。</p>
<br/>
<h5 class="header-title mt-0 mb-3">4.DeepKE开发团队</h5>
<p class="mb-0">DeepKE由浙江大学知识图谱团队协作完成，主要贡献者包括：张宁豫、陶联宽、余海洋、陈想、田玺、李磊、黎洲波、徐欣、邓淑敏、谢辛、郑国轴、陈华钧等。</p>
</div>
</div>
</div>
</div>
<div class="row" id="div_head_visualization" style="display:none;">
<div class="col-sm-12">
<div class="page-title-box">
<div class="btn-group float-right">
<ol class="breadcrumb hide-phone p-0 m-0">
<li class="breadcrumb-item active">Table</li>
<li class="breadcrumb-item active">Bar Graph</li>
<li class="breadcrumb-item active">Dynamic Graph</li>
</ol>
</div>
<h4 class="page-title">Visualization</h4>
</div>
</div>
<div class="clearfix"></div>
</div>
<div class="row" style="display:none;" id="div_visualization">
<div class="col-lg-6">
<div class="card">
<div class="card-body">
<h4 class="mt-0 header-title">Relational Triple</h4>
<p class="text-muted mb-4 font-14 " id="my_table">
The triplets are expressed in the form of a table.
</p>
</div>
</div>
</div> <!-- end col -->
<div class="col-lg-6">
<div class="card">
<div class="card-body">
<h4 class="mt-0 header-title">Dynamic Graph</h4>
<div id = "div_graph" style="width: 500px;height:350px;">
</div>
</div>
</div>
</div> <!-- end col -->
</div> <!-- end row -->
<!-- <div class="row" id="div_bargraph_whole"> -->
<div class="row" style="display:none;" id="div_bargraph_whole">
<div class="col-lg-12">
<div class="card">
<div class="card-body">
<h4 class="mt-0 header-title">Confidence</h4>
<p class="text-muted mb-4 font-14 formular">
The abscissa represents the category, and the ordinate represents the probability value.
</p>
<div id="div_bargraph" style="width: 1000px;height:400px;">
</div>
</div>
</div>
</div>
</div>
</div><!-- container -->
</div> <!-- Page content Wrapper -->
</div> <!-- content -->
<footer class="footer">
访客数：<span id="busuanzi_value_site_uv"><i class="fa fa-spinner fa-spin"></i></span>
<br/>
Copyright© 2019-2021 <a href="http://openkg.cn/">OpenKG</a> All Rights Reserved
</footer>
</div>
<!-- End Right content here -->
</div>
<!-- END wrapper -->
<!-- jQuery -->
<script src="assets//js//jquery.min.js"></script>
<script src="assets//js//popper.min.js"></script>
<script src="assets//js//bootstrap.min.js"></script>
<script src="assets//js//modernizr.min.js"></script>
<script src="assets//js//detect.js"></script>
<script src="assets//js//fastclick.js"></script>
<script src="assets//js//jquery.blockUI.js"></script>
<script src="assets//js//waves.js"></script>
<script src="assets//js//jquery.nicescroll.js"></script>
<!-- 引入不蒜子计数 -->
<script async src="//busuanzi.ibruce.info/busuanzi/2.3/busuanzi.pure.mini.js"></script>
<script src="assets/plugins/jvectormap/jquery-jvectormap-2.0.2.min.js"></script>
<script src="assets/plugins/jvectormap/jquery-jvectormap-world-mill-en.js"></script>
<script src="assets/plugins/skycons/skycons.min.js"></script>
<script src="assets/plugins/fullcalendar/vanillaCalendar.js"></script>
<script src="assets/plugins/raphael/raphael-min.js"></script>
<script src="assets/plugins/morris/morris.min.js"></script>
<script src="assets/pages/dashborad.js"></script>
<!-- App js -->
<script src="assets/js/app.js"></script>
<!-- Alertify js -->
<script src="assets/plugins/alertify/js/alertify.js"></script>
<!-- <script src="assets/pages/alertify-init.js"></script> -->
<script src="assets/js/web2triples.js"></script>
<script src="assets/incubator-echarts/dist/echarts.js"></script>
<script src="https://www.gstatic.com/charts/loader.js"></script>
<!-- <script type="text/javascript" src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"> </script> -->
</body>
</html>