Upgrade to Pro — share decks privately, control downloads, hide ads and more …

文献紹介: Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfer

Yumeto Inaoka
June 20, 2018
Research
0
110

文献紹介: Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfer

2018/06/20の文献紹介で発表

Yumeto Inaoka

June 20, 2018
Tweet

More Decks by Yumeto Inaoka

See All by Yumeto Inaoka
文献紹介: Quantity doesn’t buy quality syntax with neural language models
yumeto
1
69
文献紹介: Open Domain Web Keyphrase Extraction Beyond Language Modeling
yumeto
0
94
文献紹介: Self-Supervised_Neural_Machine_Translation
yumeto
0
73
文献紹介: Comparing and Developing Tools to Measure the Readability of Domain-Specific Texts
yumeto
0
76
文献紹介: PAWS: Paraphrase Adversaries from Word Scrambling
yumeto
0
49
文献紹介: Beyond BLEU: Training Neural Machine Translation with Semantic Similarity
yumeto
0
150
文献紹介: EditNTS: An Neural Programmer-Interpreter Model for Sentence Simplification through Explicit Editing
yumeto
0
180
文献紹介: Decomposable Neural Paraphrase Generation
yumeto
0
120
文献紹介: Analyzing the Limitations of Cross-lingual Word Embedding Mappings
yumeto
0
130

Other Decks in Research

See All in Research
ICH E9 (R1) 臨床試験のための統計的原則〜中間事象に対するストラテジー
shuntaros
1
190
ベイズ能動学習による統計的実験計画~ベイズ最適化・能動的レベル集合推定の基礎と応用~
mayumichqm
0
180
日本のZ世代における自己の外見の捉え方とケアの実態に関するリサーチ（SUMMARY版）
rs125
0
1.3k
（最先端NLP2022）DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings
chanabek326
2
540
SummerCake_pdf.pdf
lyh125
0
130
Knowledge Neurons in Pretrained Transformers (for SNLP2022)
kogoro
2
460
福岡ジャズ界のために個人サービスを開発/運営して学んだこと
daichan4649
0
260
Policy-Adaptive Estimator Selection for Off-Policy Evaluation
aiueola
0
350
サイボウズのUXリサーチャーについて
cybozuinsideout
PRO
0
380
20221219TXPMedical勉強会
tadook
1
340
データ分析の進め方とニュースメディアでのデータ活用事例 / data-analysis-in-kaggle-and-news-media
upura
0
470
再現性問題　再入門
arumakan
0
110

Featured

See All Featured
What's in a price? How to price your products and services
michaelherold
233
9.7k
What's new in Ruby 2.0
geeforr
336
30k
A designer walks into a library…
pauljervisheath
198
16k
For a Future-Friendly Web
brad_frost
166
7.7k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
32
6.7k
Art Directing for the Web. Five minutes with CSS Template Areas
malarkey
196
9.8k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
22
1.7k
Visualization
eitanlees
128
12k
Clear Off the Table
cherdarchuk
79
290k
BBQ
matthewcrist
75
8.1k
We Have a Design System, Now What?
morganepeng
37
5.9k
Git: the NoSQL Database
bkeepers
PRO
418
60k

Transcript

  1. Delete, Retrieve, Generate:
 A Simple Approach to
 Sentiment and Style

    Transfer Juncen Li, Robin Jia, He He, Percy Liang. Proceedings of NAACL-HLT 2018, pages 1865–1874, 2018. จݙ঺հ ௕Ԭٕज़Պֶେֶࣗવݴޠॲཧݚڀࣨ ҴԬເਓ

  2. "CTUSBDU wײ৘ͳͲͷଐੑΛɺଐੑʹґଘ͠ͳ͍಺༰Λ
 อ࣋ͭͭ͠ม׵Λߦ͏λεΫ wֶशʹ͸ଐੑͷΈҟͳΔΑ͏ͳจϖΞΛ࢖༻͠ͳ͍ wϑϨʔζΛ%FMFUF 3FUSJFWFͯ͠ɺͦΕΒΛݩʹ
 ࠷ऴతͳग़ྗΛ(FOFSBUF͢Δ wैདྷख๏ΑΓଟ͘ͷೖྗʹ͓͍ͯจ๏త͔ͭ ద੾ͳग़ྗ͕ੜ੒͞ΕΔ͜ͱΛਓखධՁͰ֬ೝ !2

  3. *OUSPEVDUJPO w ײ৘΍ελΠϧɺ੍࣌ͷΑ͏ͳଐੑΛ੍ޚͰ͖Δ จੜ੒ʹؔ৺͕ߴ·͍ͬͯΔ w ௨ৗɺଐੑͷΈҟͳΔύϥϨϧσʔλ͸࢖༻
 Ͱ͖ͣɺଐੑ͕ϥϕϧ෇͚͞ΕͨจͷΈΛ࢖༻ w ͜Ε·Ͱʹ("/Λ༻͍ͨख๏͕ఏҊ͞Ε͍ͯΔ͕ɺ ग़ྗ͕௿඼࣭Ͱ͋Δ͜ͱ͕ਓखධՁͰ൑໌

    !3

  4. *OUSPEVDUJPO w ଐੑʹӨڹΛ༩͑Δ୯ޠ΍۟
 Ҏ֎͸΄ͱΜͲมߋͤͣ͞ʹ
 ଐੑม׵Ͱ͖Δ৔߹͕ଟ͍ w ΑΓ୯७Ͱֶश͕؆୯ͳ
 Ұ࿈ͷγεςϜΛఏҊ !4

  5. "QQSPBDI !5

  6. "QQSPBDI !6

  7. "QQSPBDI !7

  8. "QQSPBDI !8

  9. "QQSPBDI !9

  10. "QQSPBDI !10

  11. "QQSPBDI !11

  12. "QQSPBDI !12

  13. "QQSPBDI !13

  14. "QQSPBDI !14

  15. %FMFUF w ײ৘ଐੑͷ৔߹ɺlQPTJUJWFzͷ࣌ʹݶͬͯΑ͘ग़ݱ ͢ΔOHSBNͱlOFHBUJWFzͷ࣌ʹݶͬͯΑ͘ग़ݱ
 ͢ΔOHSBNΛଐੑϚʔΧͱͯ͠࡟আ w OHSBN͔ΒଐੑΛ෼ྨ͢ΔφΠʔϒϕΠζ෼ྨث ʹ͓͚ΔOHSBNͷ৚݅෇͖֬཰͕ࢦఆͷᮢ஋Λ
 ௒͑ͨࡍʹଐੑϚʔΧͱ͢Δ !15

  16. 3FUSJFWF w ͭͷ୯ޠܥྻͷڑ཭͕Ұ൪খ͍͞΋ͷΛऔΓग़͢ w ڑ཭ͷܭࢉํ๏͸ҎԼͷͭΛ࣮ݧ  5'*%'ͰॏΈ͚ͮΒΕͨ୯ޠͷॏͳΓ  DPOUFOUFNCFEEJOHTͷϢʔΫϦουڑ཭
 ˢEFMFUFޙͷจΛ3//FODPEFSʹೖྗͨ݁͠Ռ

    !16

  17. (FOFSBUF %FMFUF0OMZ w %FMFUFޙͷจͱଐੑ͔Β%FMFUFલͷจΛ෮ݩ͢Δ Α͏ʹֶशΛߦ͏ !17

  18. (FOFSBUF %FMFUF"OE3FUSJFWF w ී௨ʹֶशͤ͞Δͱɺจ΁ଐੑϚʔΧΛ݀ຒΊ͢Δ ͚ͩͷֶशʹͳͬͯ͠·͏
 ˠεϜʔδϯά͕ߦΘΕͣྲྀெʹͳΒͳ͍ w ଐੑϚʔΧ΁֬཰తʹϊΠζΛՃ͑Δ
 ˡฤूڑ཭͕ͰಉଐੑͷผϚʔΧ΁ஔ͖׵͑Δ !18

  19. &YQFSJNFOUT w :FMQϨϏϡʔɺ"NB[POϨϏϡʔͷײ৘Λ൓స w ը૾ΩϟϓγϣϯΛΑΓϩϚϯνοΫ͔ϢʔϞϥε ʹͳΔΑ͏มߋ w ैདྷख๏ɺ)VNBO3FGFSFODFɺఏҊ͢Δͭͷ
 γεςϜΛൺֱ w

    )VNBO3FGFSFODF͸.5VSLͰऩू !19

  20. %BUBTFUT w Ωϟϓγϣϯͷςετηοτ͸ࣄ࣮ͷΈͰ͋ΔͨΊ
 ଐੑϚʔΧͷ࡟আ͸ͳ͘ɺૠೖͷΈ !20

  21. &YQFSJNFOUBM%FUBJMT w EJNFOTJPOBMXPSEWFDUPST w TJOHMFMBZFS(36XJUIIJEEFOVOJUT w NBYPVUBDUJWBUJPOGVODUJPO w "EBEFMUBXJUIBNJOJCBUDITJ[FPG w

    CFBNTFBSDIXJUIBCFBNTJ[FPG !21

  22. )VNBO&WBMVBUJPO w .5VSLͰޏͬͨϫʔΧʔ͕γεςϜͷग़ྗΛධՁ w ஈ֊ͷϦοΧʔτई౓Ͱจ๏ੑɺଐੑɺ
 ҙຯͷอ࣋ΛධՁ w ·ͨ͸ͱධՁ͞Εͨ৔߹ʹग़ྗ͸੒ޭͱݟ၏͢ w ແ࡞ҝʹநग़ͨ͠αϯϓϧΛධՁ


    ʢ֤ଐੑ͝ͱʹαϯϓϧʣ !22

  23. )VNBO&WBMVBUJPO !23

  24. $PODMVTJPO w ςΩετଐੑม׵ʹ͓͍ͯैདྷͷ("/ʹΑΔख๏ ΑΓߴੑೳͳख๏ΛఏҊ w จͷଐੑʹӨڹΛ༩͑Δ۟͸ہॴతͰ͋Δ͜ͱ͕ ޮՌΛେ͖͍ͯ͘͠Δ w কདྷతʹ͸OHSBNΑΓҰൠతͳଐੑͷ֓೦Λ։ൃ ͢Δͱ༗ӹ͕ͩɺΑΓؼೲతόΠΞεΛ൐͏

    !24