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

Dynamic_Data_Selection_for_Neural_Machine_Trans...

MARUYAMA
March 28, 2019
0
120

 Dynamic_Data_Selection_for_Neural_Machine_Translation.pdf

MARUYAMA

March 28, 2019
Tweet

More Decks by MARUYAMA

See All by MARUYAMA
vampire.pdf
tmaru0204
0
150
Misspelling_Oblivious_Word_Embedding.pdf
tmaru0204
0
170
Simple_Unsupervised_Summarization_by_Contextual_Matching.pdf
tmaru0204
0
160
Controlling_Text_Complexity_in_Neural_Machine_Translation.pdf
tmaru0204
0
140
20191028_literature-review.pdf
tmaru0204
0
140
Hint-Based_Training_for_Non-Autoregressive_Machine_Translation.pdf
tmaru0204
0
120
Soft_Contextual_Data_Augmentation_for_Neural_Machine_Translation_.pdf
tmaru0204
0
150
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

Featured

See All Featured
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
251
21k
Writing Fast Ruby
sferik
628
61k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
656
59k
Designing Experiences People Love
moore
140
23k
Six Lessons from altMBA
skipperchong
27
3.6k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
27
1.9k
Build your cross-platform service in a week with App Engine
jlugia
229
18k
Building an army of robots
kneath
303
45k
Optimising Largest Contentful Paint
csswizardry
34
3.1k
jQuery: Nuts, Bolts and Bling
dougneiner
63
7.7k
Site-Speed That Sticks
csswizardry
4
420
Building a Scalable Design System with Sketch
lauravandoore
461
33k

Transcript

  1. %ZOBNJD%BUB4FMFDUJPO GPS/FVSBM.BDIJOF5SBOTMBUJPO .BSMJFTWBOEFS8FFT "SJBOOB#JTB[[B $ISJTUPG.PO[&/.-1 QBHFTr -JUFSBUVSFSFWJFX /BHBPLB6OJWFSTJUZPG5FDIOPMPHZ5BLVNJ.BSVZBNB

  2. "CTUSBDU ⾣ֶश࣌ؒΛ࡟ݮͭͭ͠ɺੑೳΛվળ ⾣/.5ͷֶशதʹɺಈతʹֶशσʔλΛมԽͤ͞Δख๏ΛఏҊ ⾣طଘͷσʔλબ୒ख๏Λ/.5ʹదԠ͠ɺޮՌΛݕূ ˠ1#.5ʹ͸ޮՌత͕ͩɺ/.5ʹ͓͍ͯ͸͋·ΓޮՌ͕ͳ͍

  3. *OUSPEVDUJPO ⾣ͭͷ࣮ݧ ɾطଘͷσʔλબ୒ख๏ 4UBUJD%BUB4FMFDUJPO ɾ/.5ʹ߹Θͤͨσʔλબ୒ख๏ %ZOBNJD%BUB4FMFDUJPO 4BNQMJOH (SBEVBMpOFUVOJOH

  4. 4UBUJD%BUB4FMFDUJPO ɾ(FOFSBM ɾ*OEPNBJO ςετσʔλͱಉ͡υϝΠϯͷσʔλ ⾣छྨͷݴޠϞσϧΛߏங *OEPNBJO (FOFSBM 4PVSDFTJEF 5BSHFUTJEF º

    ⾣ͭͷίʔύεΛར༻

  5. 4UBUJD%BUB4FMFDUJPO $ίʔύε I*OEPNBJO G(FOFSBM  Cݴޠ fTPVSDFTJEF eUBSHFUTJEF  ⾣֤ݴޠϞσϧΛ༻͍ͯɺDSPTTFOUSPQZEJ⒎FSFODF

    $&% Λܭࢉ શֶशσʔλΛ$&%ॱʹιʔτͯ͠ɺ্ҐOจରΛֶशʹར༻

  6. %ZOBNJD%BUB4FMFDUJPO ⾣4BNQMJOH $&%ʹج͍ͮͨॏΈ෇͖αϯϓϦϯά ςετσʔλͱؔ࿈ͷ͋Δσʔλ͸બ͹Ε΍͍͢

  7. %ZOBNJD%BUB4FMFDUJPO ⾣(SBEVBMpOFUVOJOH FQPDI਺ΛॏͶΔ౓ʹɺؔ࿈ͷ௿͍σʔλΛֶशσʔλ͔Βআ֎

  8. %ZOBNJD%BUB4FMFDUJPO ⾣(SBEVBMpOFUVOJOH iFQPDI໨ͷσʔλαΠζn(i) ЋSFMBUJWFTUBSUTJ[F 㱡Ћ㱡 cGcUSBJOJOHEBUBTJ[F ЌSFUFOUJPOSBUF 㱡Ќ㱡 Б/VNCFSPGDPOTFDVUJWFFQPDI TFMFDUFETVCTFUJTVTFE

    Б㱢

  9. &YQFSJNFOUTFUUJOH ⾣(FSNBOˠ&OHMJTI ⾣.BDIJOF5SBOTMBUJPO4ZTUFNT ɾ/.5&ODPEFSEFDPEFSNPEFMXJUIHMPCBMBUUFOUJPO ɾ1#.5.PTFT

  10. &YQFSJNFOUTFUUJOH ⾣%BUBTFU ɾ&.&"NFEJDBMHVJEFMJOFT ɾ.PWJFEJBMPHVFT ɾ5&%UBMLT ɾ8.5OFXT ֶश։ൃσʔλʹ͸.JY  ධՁʹ͸֤ධՁηοτΛར༻

  11. 3FTVMU ⾣4UBUJDEBUBTFMFDUJPOGPS1#.5BOE/.5 JOEPNBJOͷσʔλͷΈ ˒  4UBUJDEBUBTFMFDUJPO ˔  JOEPNBJOͷσʔλͷΈ ˒

     4UBUJDEBUBTFMFDUJPO ˔  1#.5 /.5

  12. 3FTVMU ⾣%ZOBNJDEBUBTFMFDUJPOGPS/.5 ಉֶ͡श࣌ؒʹ͓͍ͯɺ TUBUJDTFMFDUJPOΑΓྑ͍݁Ռ 8.5OFXTΛআ͘σʔλͰ͸ɺ TUBUJDTFMFDUJPOͷ࠷ऴ݁ՌΑΓ ߴ͍είΞ

  13. 3FTVMU ⾣%ZOBNJDEBUBTFMFDUJPOGPS/.5

  14. $PODMVTJPO ⾣ֶश࣌ؒΛ࡟ݮͭͭ͠ɺੑೳΛվળ ⾣/.5ͷֶशதʹɺಈతʹֶशσʔλΛมԽͤ͞Δख๏ΛఏҊ ⾣طଘͷσʔλબ୒ख๏Λ/.5ʹదԠ͠ɺޮՌΛݕূ ˠ1#.5ʹ͸ޮՌత͕ͩɺ/.5ʹ͓͍ͯ͸͋·ΓޮՌ͕ͳ͍