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

vampire.pdf

A3ea3bc5dde6ae2dd6eae71da9c418b0?s=47 MARUYAMA
February 25, 2020
0
69

 vampire.pdf

A3ea3bc5dde6ae2dd6eae71da9c418b0?s=128

MARUYAMA

February 25, 2020
Tweet

More Decks by MARUYAMA

See All by MARUYAMA
Misspelling_Oblivious_Word_Embedding.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
52
Simple_Unsupervised_Summarization_by_Contextual_Matching.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
62
Controlling_Text_Complexity_in_Neural_Machine_Translation.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
61
20191028_literature-review.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
63
Hint-Based_Training_for_Non-Autoregressive_Machine_Translation.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
46
Soft_Contextual_Data_Augmentation_for_Neural_Machine_Translation_.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
61
An_Embarrassingly_Simple_Approach_for_Transfer_Learning_from_Pretrained_Language_Models_.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
56
Addressing_Trobulesome_Words_in_Neural_Machine_Translation.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
48
Simple_Unsupervised_Keyphrase_Extraction_using_Sentence_Embeddings.pdf
A3ea3bc5dde6ae2dd6eae71da9c418b0?s=48 tmaru0204
0
85

Featured

See All Featured
Easily Structure & Communicate Ideas using Wireframe
0c6fd6a08d0f898159cda7cafcacf07f?s=48 afnizarnur
181
15k
Support Driven Design
1519587b1a0febb658c36c94f6272b0d?s=48 roundedbygravity
86
8.5k
Clear Off the Table
Ef32e0b1ac5082450dc8c1fca3a26b5c?s=48 cherdarchuk
79
280k
What's in a price? How to price your products and services
Dad095ea7038f89f760419ce475d5d14?s=48 michaelherold
229
9.3k
Side Projects
027d1ebf66cc039a0bd3b55eeadbe75d?s=48 sachag
449
37k
Typedesign – Prime Four
A0517b08532aec8ab2aae3f65d40ad86?s=48 hannesfritz
33
1.3k
Building a Scalable Design System with Sketch
1177e050db6bafe62885362edf6e3537?s=48 lauravandoore
447
30k
GitHub's CSS Performance
64827b31aef81498662e8af4bd6d5157?s=48 jonrohan
1020
410k
Adopting Sorbet at Scale
2bb923c57a1fdc3a2eb484bb8d565fd2?s=48 ufuk
63
7.5k
Streamline your AJAX requests with AmplifyJS and jQuery
9cde37f47e4a800ea081ea42de8d749a?s=48 dougneiner
125
8.5k
A Modern Web Designer's Workflow
8081b26e05bb4354f7d65ffc34cbbd67?s=48 chriscoyier
689
180k
A Tale of Four Properties
8081b26e05bb4354f7d65ffc34cbbd67?s=48 chriscoyier
149
20k

Transcript

  1. Variational Pretraining for Semi-supervised Text Classification จݙ঺հ Suchin Gururangan, Tam

    Dang, Dallas Card, Noah A. Smith ACL2019, pages 5880–5894

  2. Abstract ⾣ܰྔͳࣄલֶशख๏ΛఏҊ ɾখن໛σʔλͰޮ཰తʹֶशՄೳ ɾߴ଎ʹಈ࡞ ⾣ςΩετ෼ྨλεΫʹ͓͍ͯɺ ɹ&-.P #&35ʹඖఢ͢ΔੑೳΛୡ੒

  3. Introduction ⾣&-.P #&35౳ͷࣄલֶशϞσϧ͕ ɹ޿͘ར༻͞Ε͍ͯΔ ⾣େن໛ͳίʔύεɾܭࢉࢿݯ͕ඞཁ ⾣ܰྔ͔ͭޮՌతͳࣄલֶशํ๏ΛఏҊ

  4. Model ⾣1SFUSBJOJOH ɾ7BSJBUJPOBM"VUP&ODPEFS

  5. Model ⾣1SFUSBJOJOH ɾ7BSJBUJPOBM"VUP&ODPEFS

  6. Model ⾣5FYUDMBTTJpDBUJPO

  7. Model ⾣5FYUDMBTTJpDBUJPO ɾ7".1*3&FNCFEEJOH݁߹ ɾ%FFQ"WFSBHF/FUXPSLͰ Τϯίʔυ

  8. Experimental setup ⾣%BUB ɾࣄલֶशσʔλ ɾ෼ྨثֶशςετ  d FYBNQMFT JOEPNBJO d

    FYBNQMFT

  9. Experimental setup ⾣-PXSFTPVSDFTFUUJOH ɾ#BTFMJOFڭࢣσʔλͷΈͰֶश ɾ4FMGUSBJOJOH൒ڭࢣ͋Γֶश  ڭࢣσʔλ Ϟσϧͷ༧ଌ݁Ռ ɾ(-07& *%

    JOEPNBJOσʔλͰֶश ɾ(-07& 0% .XPSETͷίʔύεͰֶश

  10. Results ⾣-PXSFTPVSDFTFUUJOH

  11. Results ⾣-PXSFTPVSDFTFUUJOH

  12. Experimental setup ⾣)JHISFTPVSDFTFUUJOH ɾ5SBOTGPSNFSCBTFE&-.P ɾ#&35 GSP[FO pOFUVOJOH QSFUSBJOJOH JOEPNBJO GSP[FO

    pOFUVOJOH

  13. Results ⾣)JHISFTPVSDFTFUUJOH

  14. Results ⾣)JHISFTPVSDFTFUUJOH

  15. Results ⾣$PNQVUBUJPOBM3FRVJSFNFOUT

  16. Conclusions ⾣ܰྔͳࣄલֶशख๏ΛఏҊ ɾখن໛σʔλͰޮ཰తʹֶशՄೳ ɾߴ଎ʹಈ࡞ ⾣ςΩετ෼ྨλεΫʹ͓͍ͯɺ ɹ&-.P #&35ʹඖఢ͢ΔੑೳΛୡ੒