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

Hint-Based_Training_for_Non-Autoregressive_Mach...

MARUYAMA
September 22, 2019
0
120

 Hint-Based_Training_for_Non-Autoregressive_Machine_Translation.pdf

MARUYAMA

September 22, 2019
Tweet

More Decks by MARUYAMA

See All by MARUYAMA
vampire.pdf
tmaru0204
0
140
Misspelling_Oblivious_Word_Embedding.pdf
tmaru0204
0
170
Simple_Unsupervised_Summarization_by_Contextual_Matching.pdf
tmaru0204
0
150
Controlling_Text_Complexity_in_Neural_Machine_Translation.pdf
tmaru0204
0
140
20191028_literature-review.pdf
tmaru0204
0
130
Soft_Contextual_Data_Augmentation_for_Neural_Machine_Translation_.pdf
tmaru0204
0
140
An_Embarrassingly_Simple_Approach_for_Transfer_Learning_from_Pretrained_Language_Models_.pdf
tmaru0204
0
130
Addressing_Trobulesome_Words_in_Neural_Machine_Translation.pdf
tmaru0204
0
130
Simple_Unsupervised_Keyphrase_Extraction_using_Sentence_Embeddings.pdf
tmaru0204
0
170

Featured

See All Featured
The Language of Interfaces
destraynor
155
24k
Docker and Python
trallard
43
3.2k
Designing for humans not robots
tammielis
250
25k
jQuery: Nuts, Bolts and Bling
dougneiner
62
7.6k
KATA
mclloyd
29
14k
How GitHub (no longer) Works
holman
312
140k
Building Adaptive Systems
keathley
38
2.4k
Keith and Marios Guide to Fast Websites
keithpitt
410
22k
Building Applications with DynamoDB
mza
93
6.2k
Rebuilding a faster, lazier Slack
samanthasiow
79
8.8k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
98
18k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
27
1.5k

Transcript

  1. )JOU#BTFE5SBJOJOHGPS/PO"VUPSFHSFTTJWF .BDIJOF5SBOTMBUJPO -JUFSBUVSFSFWJFX /BHBPLB6OJWFSTJUZPG5FDIOPMPHZ5BLVNJ.BSVZBNB

  2. 1BQFS ɾ&./-1*+$/-1 ɾIUUQTBSYJWPSHBCT

  3. "CTUSBDU ⾣"35ͷӅΕঢ়ଶ΍஫ऍ෼෍͔ΒzIJOUzΛಘΔֶशख๏ΛఏҊ ⾣8.5&O%F %F&Oʹ͓͍ͯ-45."35Λ্ճΔੑೳ ⾣/POBVUPSFHSFTTJWFNBDIJOFUSBOTMBUJPO /"35 ɾ"VUPSFHSFTTJWFNBDIJOFUSBOTMBUJPO "35 ʹൺ΂ͯʜ ਪ࿦࣌ͷܭࢉ࣌ؒˣ

    ຋༁ͷ඼࣭ˣ

  4. *OUSPEVDUJPO ⾣"35WT/"35 ɾ"VUPSFHSFTTJWFUSBOTMBUJPO "35 ɾ/POBVUPSFHSFTTJWFUSBOTMBUJPO /"35 "35 /"35 ݪݴޠଆจͱ໨తݴޠଆจͷฏۉจ௕ࠩ ෳ਺ͷग़ྗީิΛಘΔͨΊͷ෯

    ͜͜Ͱ͸ʜ

  5. *OUSPEVDUJPO ⾣"35ͱ/"35ͷҧ͍Λௐࠪ ɾ/"35͕ࢧ཭໓྾ͳग़ྗΛ͍ͯ͠ΔҐஔͰ͸ɺӅΕঢ়ଶͷྨࣅ౓͕ ඇৗʹߴ͘ͳΔ ɾ/"35ͷ஫ऍ෼෍͸ɺ"35ΑΓ΋ᐆດੑΛؚΉ ⾣্هͷௐࠪ݁ՌΛ΋ͱʹ/"35NPEFMΛվળ

  6. 0CTFSWBUJPO ⾣$PTJOFTJNJMBSJUJFTCFUXFFOEFDPEFSIJEEFOTUBUFT /"35Ͱ͸ɺӅΕঢ়ଶͷ ྨࣅ౓͕ඇৗʹߴ͘ɺ ಉ͡τʔΫϯΛ ग़ྗ͢Δ܏޲ʹ͋Δ

  7. 0CTFSWBUJPO ⾣"UUFOUJPOEJTUSJCVUJPO "35Ͱ͸ɺશମΛΧόʔ Ͱ͖͍ͯΔͷʹର͠ɺ /"35Ͱ͸ɺzGBSNz౳ ܽམ͍ͯ͠Δ

  8. .FUIPE ⾣)JOUTGSPNUIF"35UFBDIFS.PEFM ɾ"35 UFBDIFS Λར༻ͭͭ͠ɺ/"35 TUVEFOU Λֶश ɾ/"35ͱ"35ͷӅΕঢ়ଶ ஫ऍ෼෍ͷࠩ෼Λଛࣦؔ਺ʹ௥Ճ )JOUTGSPNXPSEBMJHONFOU

    )JOUTGSPNIJEEFOTUBUFT /FHBUJWFMPHMJLFMJIPPE

  9. .FUIPE ⾣)JOUTGSPNUIF"35UFBDIFS.PEFM ɾ)JOUTGSPNIJEEFOTUBUFT EFDPEFSͷl૚ t൪໨ͷӅΕঢ়ଶ  USUFBDIFS TUTUVEFOU U൪໨ͱU ൪໨ͷӅΕঢ়ଶ

    ͷྨࣅ౓͕ɺ ɹɾUFBDIFSଆͰ௿͘  ɹɾTUVEFOUଆͰߴ͍ ৔߹ʹϖφϧςΟΛ༩͑Δ

  10. .FUIPE ⾣)JOUTGSPNUIF"35UFBDIFS.PEFM ɾ)JOUTGSPNXPSEBMJHONFOU l૚ hϔου, t൪໨ͷ஫ऍ෼෍  USUFBDIFS TUTUVEFOU UFBDIFSͱTUVEFOUؒͷ

    ஫ऍ෼෍ͷࠩ ,-EJWFSHFODF

  11. &YQFSJNFOUT

  12. &YQFSJNFOUT ⾣"CMBUJPOTUVEJFT

  13. $PODMVTJPO ⾣"35ͷӅΕঢ়ଶ΍஫ऍ෼෍͔ΒzIJOUzΛಘΔֶशख๏ΛఏҊ ⾣8.5&O%F %F&Oʹ͓͍ͯ-45."35Λ্ճΔੑೳ ⾣/POBVUPSFHSFTTJWFNBDIJOFUSBOTMBUJPO /"35 ɾ໨ඪ຋༁ͷ඼࣭Λҡ࣋ͭͭ͠ɺਪ࿦࣌ͷܭࢉ࣌ؒΛߴ଎Խ

  14. None
  15. None