Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Multimodal Grounding for Language Processing
Search
onizuka laboratory
October 17, 2018
Research
0
51
Multimodal Grounding for Language Processing
弊研究室で行なったCOLING2018読み会の発表資料です。
onizuka laboratory
October 17, 2018
Tweet
Share
More Decks by onizuka laboratory
See All by onizuka laboratory
Phrase-Based & Neural Unsupervised Machine Translation
onilab
0
110
Tell-and-Answer: Towards Explainable Visual Question Answering using Attributes and Captions
onilab
0
71
Card-660: A Reliable Evaluation Framework for Rare Word Representation Models
onilab
0
33
A Word-Complexity Lexicon and A Neural Readability Ranking Model for Lexical Simplification
onilab
0
120
Integrating Transformer and Paraphrase Rules for Sentence Simplification
onilab
0
59
An Auto-Encoder Matching Model for Learning Utterance-Level Semantic Dependency in Dialogue Generation
onilab
0
55
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints
onilab
0
100
Modeling Multi-turn Conversation with Deep Utterance Aggregation
onilab
0
95
Learning Semantic Sentence Embeddings using Pair-wise Discriminator
onilab
0
120
Other Decks in Research
See All in Research
GPUを利用したStein Particle Filterによる点群6自由度モンテカルロSLAM
takuminakao
0
260
生成的推薦の人気バイアスの分析:暗記の観点から / JSAI2025
upura
0
260
MetaEarth: A Generative Foundation Model for Global-Scale Remote Sensing Image Generation
satai
4
190
SSII2025 [SS2] 横浜DeNAベイスターズの躍進を支えたAIプロダクト
ssii
PRO
7
4k
最適決定木を用いた処方的価格最適化
mickey_kubo
4
1.9k
データxデジタルマップで拓く ミラノ発・地域共創最前線
mapconcierge4agu
0
210
国際論文を出そう!ICRA / IROS / RA-L への論文投稿の心構えとノウハウ / RSJ2025 Luncheon Seminar
koide3
6
4.6k
20250605_新交通システム推進議連_熊本都市圏「車1割削減、渋滞半減、公共交通2倍」から考える地方都市交通政策
trafficbrain
0
770
Cross-Media Information Spaces and Architectures
signer
PRO
0
240
論文読み会 SNLP2025 Learning Dynamics of LLM Finetuning. In: ICLR 2025
s_mizuki_nlp
0
200
SNLP2025:Can Language Models Reason about Individualistic Human Values and Preferences?
yukizenimoto
0
120
大規模な2値整数計画問題に対する 効率的な重み付き局所探索法
mickey_kubo
1
360
Featured
See All Featured
RailsConf 2023
tenderlove
30
1.2k
The Power of CSS Pseudo Elements
geoffreycrofte
77
6k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
131
19k
Optimising Largest Contentful Paint
csswizardry
37
3.4k
The Cost Of JavaScript in 2023
addyosmani
53
8.9k
A designer walks into a library…
pauljervisheath
207
24k
Fight the Zombie Pattern Library - RWD Summit 2016
marcelosomers
234
17k
Site-Speed That Sticks
csswizardry
10
810
A Tale of Four Properties
chriscoyier
160
23k
A Modern Web Designer's Workflow
chriscoyier
696
190k
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
18
1.1k
ReactJS: Keep Simple. Everything can be a component!
pedronauck
667
120k
Transcript
Multimodal Grounding for Language Processing 20181017 NomotoEriko 1
### # &0 NLP +$ 65#5- Contents: Ø"!
Ø # &0 '8 Ø # 3/*2 Ø # NLP Multimodal Grounding for Language Processing 2 %7. 4( )35- ,1
Multimodal Grounding for Language Processing 3
HF>T NLP A>O… "/(# '*!.@dQaBX g6O _QR>O… KIEWLS 8e^?2JO
Conceptual grounding: ]`QZ3fDZb7P14N: U ↑ S<),%+/$,M=0 ]` [c 4 !.&- YZC;hVZC;h\ZC;h9ZC;GDZ56M=
Multimodal Grounding for Language Processing 5
# 3 - ($ '+ . ,! &" 3
- ,)* (Cross-modal transfer) ,'+ (Cross-modal interpretation) %# (Joint multimodal processing) 6 #-
@>? (Cross-modal transfer) E*D8F &E*A):F -+ 1! 92
C "G Ø07E;=→4# '45F ØB,(.E→B6F ØE< /#→F 7 $3%C
GAF (Cross-modal interpretation) $)=( /,>3D+ G@6 . K &)B"IL?*B"142;<M !N
Ø9#80LHE 5:M Ø LJ&C-5:M 8 %7'K
;$ 3 (Joint multimodal processing) =- %/ F *'
#27 : $(> I Ø+*CDG6B<* "4H Ø@A7E&)8G5? 6B 3H 9 3 !F 09.1,
Multimodal Grounding for Language Processing 10
,(+ 3 0 ,( $ * 3
.- Ø%#,( '" (Concept representations) Ø !'" (Projection) Ø&/),( '" (Compositional representations) 11 ,(+
.)50/( (Concept representations) .)50 7&;.)9 #, '!( Ø<:.)50 6,:1 +50
'!( Ø 2 8 *"3 %- 12 50$4
,*:21) (Concept representations) (&!+.F Ø"6BA7C%#/D Ø↑ “cat” ;<5 “panther”
>@5 “dog” : E 8$ : Ø04 = 3 ?=- 13 :2'9
2394 (Projection) 86>:4 D 5. &(" 86 +$) A &("@0-
(Mapping) /?>:<@ (Joint representation space) 7;86>:4 B*'!#1= ;C 14 %) &,)>:1= ←Mapping ↓ Joint learning
./40 (Projection) !"=)( (Mapping) Ø!"+!" )( !: #$% → #$'
-9 Ø87>13 @6(*? '& %$2<,: maximize sim(#$% , #$' ), minimize sim(#$% , ./0123$' ) 15 #!%#;5-9
-/70 (Projection) &@?8;B (Joint representation space) Ø%#$ :9C2+35A( ü) ?8'6<E1.
F üB 5A =*E1.F 4D 2 B !,> 16 " ?8,>
;NB KA @5 OE*KA6L/- E* I-6L CD=P U$#
< 8>7: Ø( E* RFS3.90&TJ Ø," MG? E*'U 94 Q2 )+% 17 !KA1H
NLP Multimodal Grounding for Language Processing
18
NLP #&$'&,> 957.A=<4 ØD(6' (simLex-999 etc…) 1 7.A=@2 B*
Ø/0E #&$'&7.A=%!$'&- +)?95 8C?95 :; Ø#&$'&7.A=<4" Ø#&$'&7.A=8C?953 19 #&$'& NLP
NLP !%"'% +1 3,47 Ø0) 3, 6/.( Ø0)-# $&
' 2)-,5 ' (imSitu) *8 20 !%"'% NLP imSitu*- (http://imsitu.org/ )
+. )&$ Ø " >< !,% #&$ Ø!,% #'(%
" >< - % Ø* &$ / 21 NLP
Multimodal Grounding for Language Processing 22
NLP NLP " # !
NLP 23