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
B4勉強会 書き手の判別
Search
Atsushi
February 27, 2018
94
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
B4勉強会 書き手の判別
B4勉強会
発表日 2月27日
Atsushi
February 27, 2018
More Decks by Atsushi
See All by Atsushi
文献紹介:Automated Evaluation of Out-of-Context Errors
atsumikan
0
110
文献紹介:Correction of OCR Word Segmentation Errors in Articles from the ACL Collection through Neural Machine Translation Methods
atsumikan
0
180
文献紹介:Auxiliary Objectives for Neural Error Detection Models
atsumikan
0
100
文献紹介:Wronging a Right: Generating Better Errors to Improve Grammatical Error Detection
atsumikan
0
130
文献紹介:Low-resource OCR error detection and correction in French Clinical Texts
atsumikan
0
140
文献紹介:CMMC-BDRC Solution to the NLP-TEA-2018 Chinese Grammatical Error Diagnosis Task
atsumikan
0
140
文献紹介 : Fluency Boost Learning and Inference for Neural Grammatical Error Correction
atsumikan
0
200
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
170
文献紹介:The Effect of Error Rate in Artificially Generated Data for Automatic Preposition and Determiner Correction
atsumikan
0
150
Featured
See All Featured
Sharpening the Axe: The Primacy of Toolmaking
bcantrill
46
2.9k
Dominate Local Search Results - an insider guide to GBP, reviews, and Local SEO
greggifford
PRO
0
300
What’s in a name? Adding method to the madness
productmarketing
PRO
24
4.1k
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
25
2k
A designer walks into a library…
pauljervisheath
211
24k
Designing Powerful Visuals for Engaging Learning
tmiket
1
500
Believing is Seeing
oripsolob
1
190
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
141
35k
svc-hook: hooking system calls on ARM64 by binary rewriting
retrage
2
500
Getting science done with accelerated Python computing platforms
jacobtomlinson
2
440
Un-Boring Meetings
codingconduct
0
390
Easily Structure & Communicate Ideas using Wireframe
afnizarnur
194
17k
Transcript
B4 227 B4
1
93 1. I;)% 1= 2. ,FH4> 3. #+D:
E 1. !-@CJ(AG<" 2. '$&2/B68 4. 7?.5 7?J(0* 2
$! u .+'2"% 3 - 7#
16,515 6 16,547 4 16,810 16,946 1 8( 0 ) 16,702 5, 17,028 & 16,785 3*/ 16,245
1. N-gram 2. !
4
1. N-gram 2. !
5
$= $8 %( %: 1. 0/#;,1 2.
2&'.8>) <7- 3. <76!$# 439*5 0/#; "+?) #; 6
N-gram "(< -*-- )9 1. N-gram 2. 3+,27=. ;6 1
4) !#;6, %%;6,;6 3. ;65 $( ' 4) -,/&;6-,:08- 7
N-gram - - 8 soseki_eijitsu.txt soseki_yume.txt soseki_garasu.txt soseki_omoidasu.txt ogai_niwatori.txt ogai_vita.txt
ogai_gan.txt ogai_kanoyoni.txt 200 300 400 500 600 700 Cluster Dendrogram hclust (*, "ward.D") dist(t(res)) Height
1. N-gram 2. !
9
-/''A(PCA) u 41<163-/'50 @=1&/)1 <?(+ u 8 89#",% 8
<1 u # !$ ;. 2 u 2* :-/'50 @>7 10 z1 = a11x1 + a21x2 + a31x3 + a41x4 z2 = a12x1 + a22x2 + a32x3 + a42x4 z3 = a13x1 + a23x2 + a33x3 + a43x4 z4 = a14x1 + a24x2 + a34x3 + a44x4
&"#*)@<68//J -12- u H:! ;B07F M u EKH:&"#*)!A M!>C
0G 1. &"#*)!A 2. EKH:=. ,9L3! A 3. 68//J!-68/?:!A 4. 68/?:#*'!(+$% 68/?:DI! 5 2!4(+$% 11
- - −100 −50 0 50 −20
0 20 40 60 Comp.1 Comp.2 ogai_gan ogai_kanoyoni ogai_niwatori ogai_vita soseki_eijitsu soseki_garasu soseki_omoidasu soseki_yume 12
- - −100 −50 0 50 −20
0 20 40 60 Comp.1 Comp.2 1 2 3 4 5 6 7 8 13
- #' u % +" u )
$ 247,249-251 u *() - . ! - 261-264 u &, 6600&, 14
-- 15 tensura_247.txt tensura_249.txt tensura_250.txt tensura_251.txt mushoku_264.txt mushoku_261.txt mushoku_262.txt
mushoku_263.txt 150 200 250 300 350 Cluster Dendrogram hclust (*, "ward.D") dist(t(res)) Height
-- −30 −20 −10 0 10 20 30 −10
−5 0 5 10 15 20 Comp.1 Comp.2 mushoku_261 mushoku_262 mushoku_263 mushoku_264 tensura_247 tensura_249 tensura_250 tensura_251 16
u 3. u ! u N-gram*$> u 6?:(-/+
%'> u 8 u 9" A17 @&= 4402 u #5, ;< 202 44) 17
/*,2 u -$() (2008) R &0, 1'"3+.!# page
143-153 u %4 , <https://www.macromill.com/service/data_analysis/cluster- analysis.html>,2018/02/27 18