forked from Dustzx/Typora
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path词汇.txt
425 lines (394 loc) · 18.1 KB
/
词汇.txt
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
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
ChatDoctor
been tailored to
However, such language models have not been tailored to the medical domain, resulting in poor answer accuracy and inability to give plausible recommendations for medical diagnosis, medications, etc.
---------------
SEQUENCE LEVEL TRAINING WITH RECURRENT NEURAL NETWORKS
seminal 有创意的
In their seminal work
intractable 棘手的
Unfortunately, for text generation it is generally intractable to compute an oracle of the optimal target word given the words predicted so far
-----------------------------------------------
A Survey on Transformers in Reinforcement Learning
unraveled解开
the evolution of Transformers in RL has not yet been well unraveled
taxonomy 分类学
unleashing 解放
we believe the field ofTransformRL will play an important role in unleashing the potential impact of reinforcement learning
w.r.t. 关于 with reference to
Tang and Ha leverage the attention mechanism of Transformers to process sensory sequence and construct a policy that is permutation invariant w.r.t. inputs
tempting 吸引人的
To this end 出于这一目的
To this end, we explore how RL can be tackled with the tools of sequence
modeling, using a Transformer architecture to model distributions over trajectories
and repurposing beam search as a planning algorithm
----------------------------
LCCC
facilitate 促进
All the pre-trained models and the datasets are released to facilitate future research.
technical forums 技术论坛
Other researchers resort to construct high-quality dialogue datesets by crowdsourcing
resort to 求助于
evolved升级版的
ubiquitous 普遍存在的
affinities 任务公用性;类同、
--------------------------------------
GODEL
As the same indicates 顾名思义
Goal Oriented 和Task Oriented
在没有特别说明的情况下,在人机对话系统研究工作中有着相同的含义,它们只是对同一工作的不同描述。
amenable 经得起检验的
speak to证明
Our results on these diverse tasks speak to the effectiveness of our approach.
e.g.= for example
i.e.=namely
long-standing obstacle 长期存在的问题
capitalize on 利用
BlenderBot (Roller et al., 2021) contributes a variety of models that capitalize on skills including personality, empathy, and knowledge,
engaging 迷人的
engagingness 新颖度
the same fashion 方式
ad-hoc专门的
heterogeneous异质的
significance is computed using a paired 2-sided t-test against the best competitor.
empathetic共情的
snapshot快照
apprentice新手
disentangle分辨清楚
In our evaluation setup we have been able to disentangle intrinsic and extrinsic evaluation, which respectively capture Utility and Humanness
proposition观点
The evidence that extrinsic evaluation enjoys greater inter-annotator agreement and correlation with automated evaluati on metrics further supports this proposition.
overarching 首要的
well-formedness 符合语法 well-formed
paraphrasing 解述
we hypothesize that usefulness in tasks such as QA and knowledge-grounding requires getting factual information exactly right (e.g., a year or a person’s name) with little need or room for paraphrasing.
decent 得体的 合适的 indecent
A word is like this - house. "House" is a word. It is one word and it has 5 letters (1)h (2)o (3)u (4)s (5)e.
In English, the alphabets are called "letters": abcdefghijklmnopqrstuvwxyz.
And the numbers 1-9 and 0 are called "numerals": 1234567890.
And things like a stop . and a comma , and a question mark ? are called "punctuation marks".
Special marks like @#%&* are called "symbols".
And all of these things are called "characters": abcde 12345 .,?!' @#%&*
For example, " syuo_19-5-11 " has 12 characters.
And " OldGeezer " has 9 characters.
Also, in English, a Chinese character is like this 屋 . It is one word, and it is written with one symbol, so it is one character. So in Chinese, a word and a character can be the same thing.
-----------------------------
survey
are carried out 开始进行研究
As a result, a multitude of novel works on this task are carried out
two streams of research
To the best of our knowledge据我们所知
eng
there is no room for complacency although of course we have made progress.
etc/et'set r /等等
stroke 中风
--------------------------------
prompt
ushering引领
ushering in a new era of modern natural language processing
verbalizer 标签词映射
expediently 方便地
truncation truncate 截断
Fine-Grained 细粒度的
clarity 清晰易懂
backbone 主干
inconclusive 不确定的
respirator 口罩
-----------------------------------------
Dialogue survey
converse 交谈
be retrieved from 从……中检索到
retrieval-based 基于检索的
finiteness 有限性
pipeline 管道;流水线
comprehensive 全面的 outdated 过时的
fixed 固定的
Most if not all 即使不是全部,也是绝大部分
including but not limited to 包括但不仅限于
parsing 和 analysis区别?
picture
to provide readers with a picture from the model’s perspective
rather helpful 更有帮助
Hierarchical /ˌhaɪəˈrɑːkɪk(ə)l/等级制的
discrete 各自的
inductive 归纳的
mitigates 缓和
intermediate 中间的
hippocampus 海马体
On top of that 除此之外
reuse 再利用
rather unsatisfying 相当不满意的
indicated that they were limited which made communicating with them a rather unsatisfying experience for human beings.
consensual and generic responses 万能回复
engaging 吸引人的
ubiquitous 无处不在的
wherein 以什么方式
A similar effect can be obtained by augmenting the training dataset with copies of the dialogs wherein the personality sentences are shuffled
scalability 可扩展性
interrogative 疑问的
terminologies 术语
fuse 结合 fusion
applied a multi-class classifier to flexibly fuse three distributions
disentanglement 解纠缠
to which
The pointer module pointed to the ancestor message to which the current message replies and a classifier predicted whether two messages belonged to the same thread
indistinguishable 难以区分的
simulator 模拟器;模拟者
In a nutshell 简而言之
Comparatively相较地
intercorrelated 相互关联
modular 模块化的
calibrates 矫正
This module iteratively calibrates the dialogue states based on the current input and dialogue history
ground-breaking 开创性的
lay emphasis on
Zhang et al. (2019a) argued that the hierarchical encoder-decoder does not lay enough emphasis on certain parts
granularity 粒度
discrete latent variables 离散的隐变量
treated the dialogue states as discrete latent variables to reduce the reliance on turn-level DST labels
a handful of 少量的(一小撮;一把)
-----------------------------------
UNIDS
aforementioned前面提及的
compatible 兼容的 competent胜任的
compatible for both chit-chat and taskoriented dialogues
comparably 相当地
UniDS works comparably well as the pure chit-chat system
design->devise
Different methods are devised for these two types of dialogue systems separately.
Then ->Afterwards之后
justification=reason
For fair comparisons
w.r.t. = with respect to 关于
be acquired by
Specifically, UBAR is acquired by fine-tuning the large pretrained unidirectional language model GPT-2 on the sequence of the entire dialog session which is composed of user utterance, belief state, database result, system act, and system response of every dialog turn.
------------------------------------------
UBAR
consecutive连续的
detrimental有害的
amend v.改进n.补偿
This encourages UBAR to adaptively supplement and make amends in response to the current user utterance
undergoes 经历
it first undergoes pre-training on a large number of out-ofdomain dialog turns
delexical 虚化的
catastrophic 灾难性的
underscores 强调
the big gap between few-shot BM and UBAR trained on all domains underscores the data hungry nature of end-to-end task-oriented dialog modeling
incoming 进来的
It makes sense to keep track of the belief states and add in or make adjustment to them given incoming user utterance
understudy 替补
unidirectional单向的
heuristic 启发式的
decouple 使分离
bracketed 被纳入到{}中
---------------------------
ConvLab-2
in-depth彻底的
topology 拓扑学
asynchronously 异步地
--------------------------------------
ReMeDi
coarse-grained 粗粒度的
main-subordinate 主从结构的
pseudo 假的
to facilitate future research
follow-up 后续的
overlooked 忽略
decompose 分解
Recent work considers MDSs as a kind of TDS by decomposing a MDS system into well-known sub-tasks, e.g., natural language understanding (NLU) , dialogue policy learning (DPL) , and natural language generation (NLG).
assurance 保证;保险
delexicalization 虚词化;去词汇化
monolithic 整体的
implications for 影响
The resources released with this work have broader implications in that
underexplored 未充分开发的
-------------------------------------
AIdoctor
be characterized by 具有了……特点
have been characterized by a diverse range of architectures and objectives
proliferation (数量的)激增
This limits the proliferation of scientific progress to real-world systems
modality 形式
on-ramp 入口闸道,此处为切入点
We offer practical suggestions for addressing these as an on-ramp for future work.
synthesis of prior work 综合
holistic 全面的
delve into 钻进
Although they provide an excellent holistic portrait of the subfield, they do not delve into aspects of particular interest in healthcare settings
breading 提升(添彩)
used a wider search breading their ability to provide extensive technical depth
screened 筛选过的
which were then screened using a more thorough taxonomy as part of the analysis
in concert with 与……相呼应
premier 最重要的
in-house 内部的
abbreviation 缩写词
compliance with 遵守
Questions or uncertainties regarding a paper’s compliance with inclusion criteria
facets 方面
Task-oriented dialogue systems can potentially impact many facets of healthcare in society
a close analogue for A与B很相似
systems in the cognitive or developmental health domain were a close analogue for individuals with conditions impacting memory, executive, or other cognitive function
were targeted towards 面向
Systems in the physical health domain were targeted towards individuals
closely aligned with 紧密联系
Some system objectives may be closely aligned with a single domain, whereas others may occur in many different domains
adoption or adherence 采纳
with respect to 关于
adoption or adherence with respect to healthcare outcomes
be of particular interests 特别有意义
settings背景
Engagement may be of particular interest in healthcare settings
Target Audience 目标受众
caregivers 护工
clinician 临床医师
conduct实施
Most general-domain dialogue systems research has been conducted in English
pronounced 明显的
This is even more pronounced in the healthcare domain
qualitatively and quantitatively 质量上
We examined the means through which systems were evaluated both qualitatively and quantitatively.
architectural homogeneity 架构上的同质化
This architectural homogeneity limits our understanding of good design practice within this domain
obfuscated 模糊的
exacerbate 恶化
in parallel with 同时
to date 到目前为止
We recommend further experimentation with architectural design, in parallel with work towards developing high-quality healthcare dialogue datasets, which to date remain scarce
impede 阻碍- hinder
adequate 充足的
safeguard against 预防
therapeutic stress 医疗压力
Telemedicine 远程医疗/在线医疗
gastroenterology 肠胃病学
respiratory 呼吸系统
dermatology 皮肤学
colloquial 口语的;非正式的
outpatient 门诊服务
Intensive Care Units ICU
The scenario of the former is outpatient procedure while the data from the latter is generated from Intensive Care Units (ICU).
terminology 术语
data-deficient 数据缺乏的
first attempt to 首次尝试
In this paper, we make the first attempt to develop dialog generation models about COVID-19
synthetic 合成的
real and synthetic data
on top of = in the wake of 紧接着
underserved 缺医少药的
especially to those who are underserved medically
alluring 诱人的
due to its promising potentials and
alluring commercial values, research about task oriented dialogue system (DS) has attracted increasing attention in different domains
derived from 来源于
pertinent 相关的
instantiate 列举
two-fold 分两部分的
Contributions are two-fold:
-----------------------------------
Task-oriented Dialogue System for Automatic Diagnosis
is factored into 被分为了
In practice, the user state su is factored into an agenda A (Schatzmann et al., 2007) and a goal G, noted as su = (A,G)
----------------------------------------
Inquire and Diagnose
anatomical models结构化模型
monolithic model单一构件的模型
ensemble 集成式(结构化多模块的) markdely->very
our ensemble approach obtains markedly higher disease-prediction accuracy
----------------------------------------
InstructGPT
encompasses 包括
This encompasses both explicit intentions such as following instructions and implicit intentions such as staying truthful, and not being biased, toxic, or otherwise harmful.
hallucination幻觉
disproportionately不成比例地 minority 少数的
For example, when generating text that disproportionately affects a minority group, we may want the preferences of labelers belonging to that group to be weighted more heavily.
steerability 导向性
Getting models to do what we want is directly related to the steerability and controllability literature
----------------------------------------------
Task-oriented Dialogue System for Automatic Diagnosis
cumulative 积累的
Correct symptom query and disease diagnosis will be rewarded, and the policy is learned by maximizing the expected cumulative rewards.
introspective内省的
monolithic 整体的
It shows that a particular disease is usually related to a certain group of symptoms.
triage分诊
conform 符合
conforms to the problemsolving logic for disease diagnosis in the real environment
inter alia尤其是
HRL has been successfully applied to different scenarios, inter alia, course recommendation , visual dialog , relation extraction and composite tasks with slot constraints
In the SymCat database, each disease is linked with a set of symptoms, where each symptom has a probability indicating how likely the symptom is identified for the disease
-------------------------------
DIALMED
generic 一般的 colloquial非正式的、口语的
we normalize the medications since
there are many generic names, trade names, or colloquial expressions for the same drug in dialogues
fragmentary碎片化的
outpatient procedure门诊流程
hereinafter 在下文中
consecutive 连续的
comprehensively 全面地
Secondly, our method comprehensively captures the structural and semantic information of QA pairs, which is key to understanding conversations
aggregate 整合
GAT is employed to automatically aggregate semantic and structure features on QA Dialogue Graph
relatedness关联性
prescribe 开(药)
j is prescribed by the doctor in dialogue i
noting注意
In addition, it is worth noting that the performance varies over three departments
workload工作量
---------------------------
DxFormer
upper bound 上限
Although these systems have achieved considerable diagnostic accuracy, they are still far below its performance upper bound due to few turns of interaction with patients and insufficient performance of symptom inquiry
on par with 与……相同
Considering the performance of these baseline models on the
MZ-10 dataset, the diagnostic accuracy is only on par with the lower bound of the SVM classifier (see Table 1), which suggests that these systems approximately degenerate into disease classifiers that are very weak at finding implicit symptoms.
------------------
Diaformer
unidirectional单向的
Schematic概要
the symptom recall determines the integrity of the symptom features完整性
discrete objective离散目标
gpt-4
dotted diagonal line 虚线对角线
----------------------------------
is borrowed to 用来
Despite the various designs, here we discuss and analyse the approaches under a more consistent and accessible term ‘delta-tuning’, where ‘delta’ a mathematical notation often used to
denote changes, is borrowed to refer to the portion of parameters that are ‘changed’ during training.
has ushered in 带来了
Subsequently, the advancement in pre-trained language models (PLMs) using deep transformers as their foundation has ushered in a new era of NLP
unprecedented 前所未有的
overtake 赶上
Although in-context learning has shown promising performance for PLMs such as GPT-3, fine-tuning still overtakes it under the task-specific setting.
predominant approach 主流的方法
spawn 催生出
The empirical evidence takes us one step closer to understanding how pre-trained models work and may even spawn
new theoretical questions that are worth exploring.
explore 探讨
In terms of empirical studies, we carry out extensive and systematic experiments (Results) on over 100 NLP tasks to rigorously explore the performances, combinability, the power of scale, transferability and so on.
non-trivial performance 不凡的效果
Delta-tuning yields consistent and non-trivial performance on more than 100 NLP tasks,showing that it is an effective and lightweight alternative to conventional fine-tuning.
predominantly outperforms 压倒性地胜于
……,highlighting the promising practical value of adapting super-large PLMs.
体现了调节大规模预训练语言模型的实验前景
-----------------------------
medical-related
ischemic 缺血性的
respiratory system 呼吸系统
endocrine system 内分泌系统
pediatric department 儿科
allergic rhinitis 过敏性鼻炎
rhinitis 鼻炎
pneumonia 肺炎
pre-examination 预检 triage 分诊
gastroenterology 肠胃病学
dermatology 皮肤病学
bronchi支气管 bronchitis支气管炎
heart failure 心力衰竭
upper respiratory infection 上呼吸道感染
dyspepsia消化不良
pediatric/infantile diarrhea 幼儿腹泻
type 2 diabetes 二型糖尿病
autism spectrum disorders 孤独症谱系障碍
diarrhea 腹泻
合成
synthesis of properly empathetic speech
AGI即Artificial general intelligence 通用人工智能