Upgrade to PRO for Only $50/Year—Limited-Time Offer! 🔥

文献紹介: Query and Output: Generating Words by Que...

Avatar for Yumeto Inaoka Yumeto Inaoka
July 20, 2018
Research
1
200

文献紹介: Query and Output: Generating Words by Querying Distributed Word Representations for Paraphrase Generation

2018/07/20の文献紹介で発表

Avatar for Yumeto Inaoka

Yumeto Inaoka

July 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
Avatar for Yumeto Inaoka yumeto
1
200
文献紹介: Open Domain Web Keyphrase Extraction Beyond Language Modeling
Avatar for Yumeto Inaoka yumeto
0
250
文献紹介: Self-Supervised_Neural_Machine_Translation
Avatar for Yumeto Inaoka yumeto
0
170
文献紹介: Comparing and Developing Tools to Measure the Readability of Domain-Specific Texts
Avatar for Yumeto Inaoka yumeto
0
180
文献紹介: PAWS: Paraphrase Adversaries from Word Scrambling
Avatar for Yumeto Inaoka yumeto
0
170
文献紹介: Beyond BLEU: Training Neural Machine Translation with Semantic Similarity
Avatar for Yumeto Inaoka yumeto
0
290
文献紹介: EditNTS: An Neural Programmer-Interpreter Model for Sentence Simplification through Explicit Editing
Avatar for Yumeto Inaoka yumeto
0
360
文献紹介: Decomposable Neural Paraphrase Generation
Avatar for Yumeto Inaoka yumeto
0
240
文献紹介: Analyzing the Limitations of Cross-lingual Word Embedding Mappings
Avatar for Yumeto Inaoka yumeto
0
240

Other Decks in Research

See All in Research
ForestCast: Forecasting Deforestation Risk at Scale with Deep Learning
Avatar for SatAI.challenge satai
2
120
POI: Proof of Identity
Avatar for MATSUMOTO Katsuyoshi katsyoshi
0
120
SREのためのテレメトリー技術の探究 / Telemetry for SRE
Avatar for Yuuki Tsubouchi (yuuk1) yuukit
13
2.6k
超高速データサイエンス
Avatar for Yusuke Matsui matsui_528
1
320
[IBIS 2025] 深層基盤モデルのための強化学習 驚きから理論にもとづく納得へ
Avatar for Akifumi Wachi akifumi_wachi
19
9k
Mamba-in-Mamba: Centralized Mamba-Cross-Scan in Tokenized Mamba Model for Hyperspectral Image Classification
Avatar for SatAI.challenge satai
3
350
論文紹介: ReGenesis: LLMs can Grow into Reasoning Generalists via Self-Improvement
Avatar for Hisao Katsumi hisaokatsumi
0
150
CoRL2025速報
Avatar for robotpaper.challenge rpc
3
3.7k
[Devfest Incheon 2025] 모두를 위한 친절한 언어모델(LLM) 학습 가이드
Avatar for Beomi beomi
2
1.3k
財務諸表監査のための逐次検定
Avatar for MasaKat0 masakat0
0
210
世界の人気アプリ100個を分析して見えたペイウォール設計の心得
Avatar for Akihiro Kokubo akihiro_kokubo
PRO
65
35k
SNLP2025:Can Language Models Reason about Individualistic Human Values and Preferences?
Avatar for Yuki Zenimoto yukizenimoto
0
230

Featured

See All Featured
The Mindset for Success: Future Career Progression
Avatar for Greg Gifford greggifford
PRO
0
200
Lessons Learnt from Crawling 1000+ Websites
Avatar for Charles Meaden charlesmeaden
0
950
The Spectacular Lies of Maps
Avatar for Per Axbom axbom
PRO
1
400
So, you think you're a good person
Avatar for Per Axbom axbom
PRO
0
1.8k
Collaborative Software Design: How to facilitate domain modelling decisions
Avatar for Kenny Baas-Schwegler baasie
0
100
Dominate Local Search Results - an insider guide to GBP, reviews, and Local SEO
Avatar for Greg Gifford greggifford
PRO
0
17
Breaking role norms: Why Content Design is so much more than writing copy - Taylor Woolridge
Avatar for UX Y'all uxyall
0
120
Self-Hosted WebAssembly Runtime for Runtime-Neutral Checkpoint/Restore in Edge–Cloud Continuum
Avatar for Yuki Nakata chikuwait chikuwait
0
230
We Have a Design System, Now What?
Avatar for Morgane Peng morganepeng
54
7.9k
Exploring anti-patterns in Rails
Avatar for Elle Meredith aemeredith
2
210
Unsuck your backbone
Avatar for Amy Palamountain ammeep
671
58k
Designing for Timeless Needs
Avatar for Cassini Nazir cassininazir
0
93

Transcript

  1. Query and Output:
 Generating Words by Querying
 Distributed Word Representations


    for Paraphrase Generation Shuming Ma, Xu Sun, Wei Li, Sujian Li, Wenjie Li, Xuancheng Ren.
 Proceedings of NAACL-HLT 2018, pages 196-206, 2018. จݙ঺հ ௕Ԭٕज़Պֶେֶࣗવݴޠॲཧݚڀࣨ ҴԬເਓ

  2. "CTUSBDU wطଘͷ4FRTFRϞσϧ͕ੜ੒͢Δจ͸
 จ๏తʹਖ਼͍͕͠ҙຯతʹෆద੾ͳ͜ͱ͕Α͋͘Δ w୯ޠ෼ࢄදݱͷݕࡧʹΑͬͯ୯ޠΛੜ੒͢Δ
 8PSE&NCFEEJOH"UUFOUJPO/FUXPSL 8&"/ 
 ΛఏҊ wݴ͍׵͕͑ॏཁͳςΩετฏқԽͱ୹จཁ໿λεΫͰ
 TUBUFPGUIFBSUΛୡ੒

    !2

  3. *OUSPEVDUJPO w ैདྷͷ4FRTFRϞσϧ͸୯ޠͷҙຯͰ͸ͳ͘܇࿅ ηοτͷ୯ޠ΍ύλʔϯΛ҉ه͢Δ܏޲͕͋Δ
 ˡσίʔμͷग़ྗ૚͕ҙຯత৘ใΛϞσϦϯά
 ɹ͍ͯ͠ͳ͍ͨΊ w σίʔμͷग़ྗ૚͕࣋ͭύϥϝʔλ਺͕ଟ͍
 ӅΕ૚ͷ࣍ݩ͕ ޠኮαΠζ͕ສͷ৔߹


    ύϥϝʔλ਺͸ ສͱͳΔ !3

  4. 8&"/ w 8PSE&NCFEEJOH"UUFOUJPO/FUXPSL w 3//ͷग़ྗΛΫΤϦͱͯ͠࠷΋Ұக͢Δ෼ࢄදݱ ͷ୯ޠΛBUUFOUJPOΛ࢖༻ͨ͠ݕࡧʹΑͬͯબ୒ w ୯ޠͷ෼ࢄදݱ͸Τϯίʔμɺσίʔμ΁ͷೖྗ ʹՃ͑ͯग़ྗ૚ͷΫΤϦʹΑͬͯ΋ߋ৽͞ΕΔ !4

  5. !5 8&"/

  6. !6 8&"/

  7. !7 8&"/

  8. !8 8&"/

  9. ΫΤϦͱ୯ޠͷηοτ͔ΒείΞΛܭࢉ RUλΠϜεςοϓUͷRVFSZ
 FJJ൪໨ͷީิ୯ޠ J  ʜ O OޠኮαΠζ
 είΞ͕࠷େͱͳΔ୯ޠΛબ୒
 WBMJEBUJPOTFUTͰͷੑೳΛجʹHFOFSBMΛ࢖༻

    !9

  10. 5SBJOJOH w ୯ޠͷબ୒ʹݕࡧΛ༻͍͍ͯΔ͕ɺҰൠతͳ 4FRTFRͱಉ༷ʹඍ෼Մೳ
 ˠଛࣦؔ਺͸ಉ͡΋ͷ͕࢖༻Մೳ w "EBN
 Ћ Ќ Ќ

    ЏF !10

  11. &YQFSJNFOUT 5FYU4JNQMJpDBUJPO w %BUBTFUT
 1BSBMMFM8JLJQFEJB4JNQMJpDBUJPO$PSQVT 18,1 
   USBJOWBMJEUFTU

    
 &OHMJTI8JLJQFEJBBOE4JNQMF&OHMJTI8JLJQFEJB
  &84&8 
 L  USBJOWBMJEUFTU 
 UFTUTFU͸".5ͰಘΒΕͨͭͷ3FGFSFODFΛ࣋ͭ !11

  12. &YQFSJNFOUT 5FYU4JNQMJpDBUJPO w &WBMVBUJPO.FUSJDT
 #-&6
 ػց຋༁΍ฏқԽͰ޿͘༻͍ΒΕ͍ͯΔࣗಈධՁख๏
 
 ਓखධՁ
 ྲྀெੑɺଥ౰ੑɺฏқ͞ΛͰධՁ
 ฏқ͞͸ग़ྗ͕ೖྗͱൺ΂ͯͲΕ͚ͩฏқ͔Λࣔ͢

    !12

  13. ݁Ռ ࣗಈධՁ !13

  14. ݁Ռ ਓखධՁ !14

  15. "OBMZTJT w 8&"/͸ैདྷͷ4FRTFRͱൺ΂ͯύϥϝʔλ͕গͳ͍ !15

  16. "OBMZTJT w /54XW΍1#.53͸ඞਢͷཁૉΛ͍͍ܽͯΔ w 4#.54"3*͸ྲྀெ͕ͩҙຯ͕ҟͳΔ !16

  17. "OBMZTJT !17

  18. "OBMZTJT !18 ˢলུ͕ଟ͘৘ใ͕ෆ଍͍ͯ͠Δ

  19. "OBMZTJT !19 ˢTJFNFOTNBSUJO SSC TIVSCBͱ͍ͬͨແؔ܎ͷ
 ɹ୯ޠΛग़ྗ

  20. "OBMZTJT !20 ˣྲྀெ͕ͩҙຯ͕ҟͳΓΑΓཧղ͕೉͘͠ͳ͍ͬͯΔ

  21. $PODMVTJPO w ΫΤϦʹΑΔ୯ޠ෼ࢄදݱͷݕࡧ͔Β୯ޠΛੜ੒ ͢ΔFODPEFSEFDPEFSGSBNFXPSLΛఏҊ w ͭͷӳޠฏқԽσʔληοτʹ͓͍ͯ
 ϕʔεϥΠϯͱൺֱͯ͠#-&6͕ͦΕͧΕ
 ͓Αͼ޲্ͨ͠ w ຊϞσϧ͸TUBUFPGUIFBSUΛୡ੒͍ͯ͠Δ

    !21