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
文献紹介 : Fluency Boost Learning and Inference for Neural Grammatical Error Correction
Search
Atsushi
August 29, 2018
0
150
文献紹介 : Fluency Boost Learning and Inference for Neural Grammatical Error Correction
2018年8月29日 文献紹介
長岡技術科学大学
自然言語処理研究室
Atsushi
August 29, 2018
Tweet
Share
More Decks by Atsushi
See All by Atsushi
文献紹介:Automated Evaluation of Out-of-Context Errors
atsumikan
0
76
文献紹介:Correction of OCR Word Segmentation Errors in Articles from the ACL Collection through Neural Machine Translation Methods
atsumikan
0
130
文献紹介:Auxiliary Objectives for Neural Error Detection Models
atsumikan
0
66
文献紹介:Wronging a Right: Generating Better Errors to Improve Grammatical Error Detection
atsumikan
0
93
文献紹介:Low-resource OCR error detection and correction in French Clinical Texts
atsumikan
0
75
文献紹介:CMMC-BDRC Solution to the NLP-TEA-2018 Chinese Grammatical Error Diagnosis Task
atsumikan
0
92
文献紹介:語彙の概念化と Wikipediaを用いた英字略語の意味推定方法
atsumikan
0
120
文献紹介:The Effect of Error Rate in Artificially Generated Data for Automatic Preposition and Determiner Correction
atsumikan
0
110
文献紹介: Automatic Annotation and Evaluation of Error Types for Grammatical Error Correction
atsumikan
0
140
Featured
See All Featured
GraphQLとの向き合い方2022年版
quramy
28
12k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
19
1.9k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
153
14k
Fontdeck: Realign not Redesign
paulrobertlloyd
75
4.8k
10 Git Anti Patterns You Should be Aware of
lemiorhan
644
57k
Art, The Web, and Tiny UX
lynnandtonic
288
19k
How to train your dragon (web standard)
notwaldorf
71
5.1k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
57
14k
Design by the Numbers
sachag
274
18k
Making the Leap to Tech Lead
cromwellryan
122
8.4k
Why You Should Never Use an ORM
jnunemaker
PRO
50
8.6k
How GitHub (no longer) Works
holman
301
140k
Transcript
'MVFODZ#PPTU-FBSOJOHBOE*OGFSFODF GPS/FVSBM(SBNNBUJDBM&SSPS$PSSFDUJPO Ԭٕज़Պֶେֶࣗવݴޠॲཧݚڀࣨ ੁ३ࢤ จݙհ ݄ 5BP(F'VSV8FJ.JOH;IPV 1SPDFFEJOHTPGUIFUI"OOVBM.FFUJOHPGUIF"TTPDJBUJPOGPS $PNQVUBUJPOBM-JOHVJTUJDT -POH1BQFST
QBHFTr .FMCPVSOF "VTUSBMJB +VMZ 1
"CTUSBDU w ଟ͘ͷ4FR4FRϞσϧʹͭͷݶք͕͋Δ w ݶఆ͞ΕͨޡΓగਖ਼σʔλͰҰൠԽͰ͖ͳ͍Մೳੑ͕͋Δ w ෳͷޡΓ͕͋ΔจΛશʹగਖ਼Ͱ͖ͳ͍͜ͱ͕͋Δ w ͭͷݶքΛղܾ͢Δख๏ΛఏҊ͢Δ w
$P/--ͱ+'-&(CFODINBSLEBUBTFUTͰ 4P5"Λୡͨ͠ !2
*OUSPEVDUJPO w ଟ͘ͷ4FR4FRϞσϧʹͭͷݶք͕͋Δ ݶΒΕͨ܇࿅σʔλΛ༻͍ͯ͠ΔͨΊ Ϟσϧ͕ҰൠԽ͍ͯ͠ͳ͍Մೳੑ͕͋Δ ෳͷޡΓΛؚΉจʹର͠ Ұͷగਖ਼Ͱ
ෳͷޡΓΛશʹగਖ਼͢Δ͜ͱ͕Ͱ͖ͳ͍ w ্هͷݶքʹऔΓΉͨΊʹ ৽͍͠ྲྀெΛߴΊΔֶशͱਪख๏ΛఏҊ͢Δ !3
'MVFODZCPPTUMFBSOJOH w จYͷྲྀெG Y !4 ݴޠϞσϧʹΑͬͯܭࢉ͞Εͨɹͷจ຺ͷ֬ จɹͷަࠩΤϯτϩϐʔ w ຊจͰ࣍ͷ͍ͣΕ͔ͷख๏ͰྲྀெΛߴΊΔ
#BDLCPPTUMFBSOJOH 4FMGCPPTUMFBSOJOH %VBMCPPTUMFBSOJOH P(xi |x<i ) : H(x) : x<i x
#BDLCPPTUMFBSOJOH w #BDLUSBOTMBUJPOΛ༻ͯ͠ྲྀெͳจ ͔ΒඇྲྀெͳจΛੜ͠τϨʔχϯά σʔλʹՃ͢Δ !5 ޡΓగਖ਼͞Εͨྲྀெͳจ ੜ͞ΕͨޡΓจͷީิ xc
: x0k :
4FMGCPPTUMFBSOJOH w ޡΓగਖ਼݁Ռͷจͷ͏ͪ ྲྀெͳจΑΓ ྲྀெ͕͍จΛΤϥʔ༧ଌͱΈͳͯ͠ τϨʔχϯάσʔλʹՃ͢Δ !6 ޡΓగਖ਼͞ΕΔલͷඇྲྀெͳจ ޡΓగਖ਼͞Εͨจͷީิ
ޡΓగਖ਼͞Εͨྲྀெͳจ xr : x0k : xc :
%VBMCPPTUMFBSOJOH !7 w #BDLCPPTUMFBSOJOHͱ 4FMGCPPTUMFBSOJOHΛ Έ߹ΘͤͨϞσϧ ޡΓగਖ਼͞ΕΔલͷ ඇྲྀெͳจ ޡΓగਖ਼͞Εͨจͷީิ
ޡΓగਖ਼͞Εͨྲྀெͳจ xr : x0k : xc :
'MVFODZCPPTUJOGFSFODF !8
%BUBTFU w %BUBTFU w $P/--TIBSFEUBTLEBUBTFU w -BOHDPN w OBUJWFEBUB&OHMJTI8JLJQFEJBDPSQVT w
-BOHVBHF.PEFM -. w MBZFS(363//MBOHVBHFNPEFM .JLPMPWFUBM !9
&YQFSJNFOUBMSFTVMUT !10 w OPSNBMTFRTFR w 3FDVSSFOU/FVSBM/FUXPSL 3// CBTFEFODPEFSEFDPEFS TFRTFRNPEFM 4VUTLFWFSFUBM
$IPFUBM
!11
!12
$PODMVTJPO w ҎલͷOFVSBM(&$ϞσϧͷݶքΛࠀ͢ΔͨΊͷ ৽͍͠ྲྀெΛߴΊΔख๏ΛఏҊͨ͠ w ޡΓͷ͋ΔσʔλͱωΠςΟϒσʔλΛશʹ׆༻͠ ௨ৗͷ4FR4FRͷֶशΑΓύϑΥʔϚϯεΛେ෯ʹվળͨ͠ w
$P/--ͱ+'-&(CFODINBSLEBUBTFUTͰ 4P5"Λୡͨ͠ !13