Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
Argument Invention from First Principles
Search
ando
September 26, 2019
Research
360
2
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Argument Invention from First Principles
最先端NLP勉強会2019
ando
September 26, 2019
More Decks by ando
See All by ando
(Reading )Does BERT Know that the IS-A Relation Is Transitive?
ando55
0
130
博士論文公聴会資料
ando55
0
540
Is In-hospital Meta-information Useful for Abstractive Discharge Summary Generation?
ando55
0
290
(読み会)Evaluating Factuality in Text Simplification
ando55
1
210
(Reading) Relational Multi-Task Learning Modeling Relations between Data and Tasks
ando55
0
200
(Reading )Agreeing to Disagree: Annotating Offensive Language Datasets with Annotators’ Disagreement
ando55
0
180
(Reading) Preregistering NLP research
ando55
0
78
(Reading) Predictive Adversarial Learning from Positive and Unlabeled Data
ando55
0
170
Other Decks in Research
See All in Research
量子サマースクール2026「量子計算機アーキテクチャ分野の概観」
youten622
1
910
第64回CV・PRML勉強会 論文紹介:Linguistic Priors for Visual Decoupling: Towards Symmetric Vision-Brain Alignment
sokikatayama
0
210
Physical AIでモデリングはどう変わるか/How Physical AI Will Change Modeling
stktu
0
260
HAKARI-Bench - 実運用視点での情報検索モデル評価ベンチマーク
hotchpotch
1
760
Anthropic が提案する LLM の内部状態を自然言語で説明可能にした Natural Language Autoencoders / Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations
shunk031
0
280
SAM3を用いたコマ・吹き出しの 領域検出と分割構造からの読み順推定
kzmssk
0
140
PGDM: Physically Guided Diffusion Model for L Downscaling
satai
3
490
実例から見るLLMのマンガ理解:実務VQAタスクによる長期的文脈と視覚情報の定性評価
kzmssk
0
140
多様なデータを許容し学習し続ける模倣学習 / Advanced Imitation Learning for VLA
prinlab
0
330
大規模言語モデルは誰を覚えているか / Who Do Large Language Models Memorize?
upura
0
190
データサイエンティストの就労意識~2015 → 2026 一般(個人)会員アンケートより
datascientistsociety
PRO
0
890
[CV勉強会@関東 CVPR2026] PSDesigner: Automated Graphic Design with a Human-Like Creative Workflow / kantocv 67th CVPR 2026
shunk031
0
350
Featured
See All Featured
Information Architects: The Missing Link in Design Systems
soysaucechin
1
1.2k
Why Your Marketing Sucks and What You Can Do About It - Sophie Logan
marketingsoph
0
410
Accessibility Awareness
sabderemane
1
220
Speed Design
sergeychernyshev
33
2.1k
Easily Structure & Communicate Ideas using Wireframe
afnizarnur
194
17k
Build your cross-platform service in a week with App Engine
jlugia
234
19k
Hiding What from Whom? A Critical Review of the History of Programming languages for Music
tomoyanonymous
3
1.3k
Fight the Zombie Pattern Library - RWD Summit 2016
marcelosomers
234
18k
Large-scale JavaScript Application Architecture
addyosmani
515
110k
Bash Introduction
62gerente
615
220k
Test your architecture with Archunit
thirion
2
2.4k
The Illustrated Children's Guide to Kubernetes
chrisshort
51
53k
Transcript
"3(6.&/5*/7&/5*0/ '30.'*34513*/$*1-&4 , , , , , D2
ACL 2019 1 ࠷ઌ/-1ษڧձ
֓ཁ • i e X o t A ) (
r : ) + + u 9 9 +1 / ,) : o t mx s t e g ln 2
"3(6.&/5.*/*/( 3 • ? - / $(&!.)/!+/:6;=2!. ':6<A70><8=:6<@ /,%( "531
70><8=:6< /429=7 $ *)/#$ ⇨ / -( " https://www.atelier-roi.co.jp/the-uses-of-argument/
Γ͍ͨ͜ͱ 4 Clean energy humanity must further exploit solar energy.
(further exploit, solar energy) KB ! claim "Claim !
ʢকདྷతʹʣΓ͍ͨ͜ͱ 5 Clean energy Humanity must embrace clean energy in
order to fight climate change (fight, global warming) KB Claim suggest Ecological concerns add further strain on the economy suggest
ఏҊख๏ɿσʔλͷߏ • ( , ) A = : C :
C () ) 6 ex: Animal rights Claim ex: ban Claimtopic ex: fast food
.05*0/ ( ( • ( , ) = • (
2 = 4 7 Appendix
"3(6.&/5 • ) (, 8 •
'"!& $% 37!&# ( • Claim ) '"!& $%(
σʔλ࡞ i 7 376 0 376 7830 sm o S
7 376C ( , - 3 o c o A - 3 o p . 2219 i ) , S - 3 7 376 S 7 376 0 3 7 376 e - 3 7 376 ln t u 0 3 7 376e g a ) 9
σʔλ֦ு ( , 357154 07154 75610 o 9 l .
13 ( , 357154 9 ( 357154 0 13 a8 A A , t 357154 357154 C C8- cC 9 8. 13 357154 mp n 0 13 357154 i ) 10
$-"*.ͷҰൠੑΛ֬ೝ͢Δ e a -2 2 Mm 5 2 a M
C 5 2 +M A c ( e A +Ml k o 800 1 + no 800 1 ,4.25 st h M r 800 1 +h i p )( 11
.05*0/ͷྨ • 0 1, C A • By action (BA-k)
motion(a,t)$CoPA(c)- &3p(c,a)5%) • By topic, nearest neighbors (KNN) motion$"243+,1&3(Dor+, 2018)motion5& • By topic, word2vec features (W2V) motion(a, t).topic5w2v*Embedding >:;=7<8 *;96 • By topic, Naive Bayes (NB) and Recurrent Neural Network (RNN) t- /43 5 0CoPA5+%)NB, RNN • By topic and action (LR): feature engineering*17.feature5(' • Ensemble 6..,#*!;96.CoPA5 12
݁Ռ • 6 8 • % % • % %
% %% 6 13
ࡾஈ๏ͷద༻ 14 Humanity must embrace clean energy in order to
fight climate change. Solar energy is a form of clean energy Therefore, humanity must further exploit solar energy. %CoPAClaim& %motionCoPAClean energy"$# & %motion! & 'α
·ͱΊɾײ l 3 El n p ms e l ms
p F A F b F o i3 1 r • 12 / / 1 E E • 2 . c 15
͓·͚ʙ 3)&503*$ʙ • • IE 16
3)&503*$ͱ • oeCE sD a u p tup •
'" $!&#! ) )(%'#! • v l I mMP i • tuc MP , r ( n v / , ) 17
ൃʢ */7&/5*0ʣ • ) n ) ) • Topics$'*%"#(*(!()&≒ &
• d k ) o pk c ) t pk • T ) i ) ) ) ) ) s ) ) () ) ) ) 18 1 , 0./-, ―$, +