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
vampire.pdf
Search
Sponsored
·
Your Podcast. Everywhere. Effortlessly.
Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
→
MARUYAMA
February 25, 2020
200
0
Share
vampire.pdf
MARUYAMA
February 25, 2020
More Decks by MARUYAMA
See All by MARUYAMA
Misspelling_Oblivious_Word_Embedding.pdf
tmaru0204
0
210
Simple_Unsupervised_Summarization_by_Contextual_Matching.pdf
tmaru0204
0
200
Controlling_Text_Complexity_in_Neural_Machine_Translation.pdf
tmaru0204
0
180
20191028_literature-review.pdf
tmaru0204
0
160
Hint-Based_Training_for_Non-Autoregressive_Machine_Translation.pdf
tmaru0204
0
150
Soft_Contextual_Data_Augmentation_for_Neural_Machine_Translation_.pdf
tmaru0204
0
180
An_Embarrassingly_Simple_Approach_for_Transfer_Learning_from_Pretrained_Language_Models_.pdf
tmaru0204
0
160
Addressing_Trobulesome_Words_in_Neural_Machine_Translation.pdf
tmaru0204
0
180
Simple_Unsupervised_Keyphrase_Extraction_using_Sentence_Embeddings.pdf
tmaru0204
0
200
Featured
See All Featured
End of SEO as We Know It (SMX Advanced Version)
ipullrank
3
4.2k
Fireside Chat
paigeccino
42
3.9k
Navigating Team Friction
lara
192
16k
Agile Actions for Facilitating Distributed Teams - ADO2019
mkilby
0
200
The agentic SEO stack - context over prompts
schlessera
0
790
Marketing to machines
jonoalderson
1
5.3k
Beyond borders and beyond the search box: How to win the global "messy middle" with AI-driven SEO
davidcarrasco
3
150
The Curse of the Amulet
leimatthew05
1
13k
4 Signs Your Business is Dying
shpigford
187
22k
Building Experiences: Design Systems, User Experience, and Full Site Editing
marktimemedia
0
520
Building the Perfect Custom Keyboard
takai
2
780
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
254
22k
Transcript
Variational Pretraining for Semi-supervised Text Classification จݙհ Suchin Gururangan, Tam
Dang, Dallas Card, Noah A. Smith ACL2019, pages 5880–5894
Abstract ⾣ܰྔͳࣄલֶशख๏ΛఏҊ ɾখنσʔλͰޮతʹֶशՄೳ ɾߴʹಈ࡞ ⾣ςΩετྨλεΫʹ͓͍ͯɺ ɹ&-.P #&35ʹඖఢ͢ΔੑೳΛୡ
Introduction ⾣&-.P #&35ͷࣄલֶशϞσϧ͕ ɹ͘ར༻͞Ε͍ͯΔ ⾣େنͳίʔύεɾܭࢉࢿݯ͕ඞཁ ⾣ܰྔ͔ͭޮՌతͳࣄલֶशํ๏ΛఏҊ
Model ⾣1SFUSBJOJOH ɾ7BSJBUJPOBM"VUP&ODPEFS
Model ⾣1SFUSBJOJOH ɾ7BSJBUJPOBM"VUP&ODPEFS
Model ⾣5FYUDMBTTJpDBUJPO
Model ⾣5FYUDMBTTJpDBUJPO ɾ7".1*3&FNCFEEJOH݁߹ ɾ%FFQ"WFSBHF/FUXPSLͰ Τϯίʔυ
Experimental setup ⾣%BUB ɾࣄલֶशσʔλ ɾྨثֶशςετ d FYBNQMFT JOEPNBJO d
FYBNQMFT
Experimental setup ⾣-PXSFTPVSDFTFUUJOH ɾ#BTFMJOFڭࢣσʔλͷΈͰֶश ɾ4FMGUSBJOJOHڭࢣ͋Γֶश ڭࢣσʔλ Ϟσϧͷ༧ଌ݁Ռ ɾ(-07& *%
JOEPNBJOσʔλͰֶश ɾ(-07& 0% .XPSETͷίʔύεͰֶश
Results ⾣-PXSFTPVSDFTFUUJOH
Results ⾣-PXSFTPVSDFTFUUJOH
Experimental setup ⾣)JHISFTPVSDFTFUUJOH ɾ5SBOTGPSNFSCBTFE&-.P ɾ#&35 GSP[FO pOFUVOJOH QSFUSBJOJOH JOEPNBJO GSP[FO
pOFUVOJOH
Results ⾣)JHISFTPVSDFTFUUJOH
Results ⾣)JHISFTPVSDFTFUUJOH
Results ⾣$PNQVUBUJPOBM3FRVJSFNFOUT
Conclusions ⾣ܰྔͳࣄલֶशख๏ΛఏҊ ɾখنσʔλͰޮతʹֶशՄೳ ɾߴʹಈ࡞ ⾣ςΩετྨλεΫʹ͓͍ͯɺ ɹ&-.P #&35ʹඖఢ͢ΔੑೳΛୡ