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
Spacyでお手軽NLP / NLP with spacy
Search
Sponsored
·
SiteGround - Reliable hosting with speed, security, and support you can count on.
→
himkt
June 13, 2018
Programming
1.1k
0
Share
Spacyでお手軽NLP / NLP with spacy
2018/06/13のレトリバセミナーのスライドです
himkt
June 13, 2018
More Decks by himkt
See All by himkt
Linformer: paper reading
himkt
0
620
RoBERTa: paper reading
himkt
1
410
NLP SoTA 勉強会 / ner_2019
himkt
2
1.5k
自然言語処理 @ クックパッド / nlp at cookpad
himkt
1
580
Interpretable Machine Learning 6.3 - Prototypes and Criticisms
himkt
2
200
ニューラル固有表現抽出 / Neural Named Entity Recognition
himkt
3
820
ニューラル固有表現抽出器を実装してみる / PyNER
himkt
6
2.2k
Deep Learning Book 10その2 / deep learning book 10 vol2
himkt
2
210
ふわふわ系列ラベリング / ner 2018
himkt
5
870
Other Decks in Programming
See All in Programming
AI-DLC Deep Dive
yuukiyo
9
4.8k
検索設計から 推論設計への重心移動と Recall-First Retrieval
po3rin
4
1.2k
「話せることがない」を乗り越える 〜日常業務から登壇テーマをつくる思考法〜
shoheimitani
4
860
属人化しないコード品質の作り方_2026.04.07.pdf
muraaano
0
240
PHP で mp3 プレイヤーを実装しよう
m3m0r7
PRO
0
290
事業会社でのセキュリティ長期インターンについて
masachikaura
1
270
ルールルルルルRubyの中身の予備知識 ── RubyKaigiの前に予習しなイカ?
ydah
1
210
Cache-moi si tu peux : patterns et pièges du cache en production - Devoxx France 2026 - Conférence
slecache
0
310
SkillがSkillを生む:QA観点出しを自動化した
sontixyou
6
3.5k
YJITとZJITにはイカなる違いがあるのか?
nakiym
0
250
The Less-Told Story of Socket Timeouts
coe401_
3
640
HTML-Aware ERB: The Path to Reactive Rendering @ RubyKaigi 2026, Hakodate, Japan
marcoroth
0
290
Featured
See All Featured
Tips & Tricks on How to Get Your First Job In Tech
honzajavorek
1
500
Navigating the moral maze — ethical principles for Al-driven product design
skipperchong
2
350
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
35
2.4k
Pawsitive SEO: Lessons from My Dog (and Many Mistakes) on Thriving as a Consultant in the Age of AI
davidcarrasco
0
120
Darren the Foodie - Storyboard
khoart
PRO
3
3.3k
Bridging the Design Gap: How Collaborative Modelling removes blockers to flow between stakeholders and teams @FastFlow conf
baasie
0
520
Building Experiences: Design Systems, User Experience, and Full Site Editing
marktimemedia
0
490
How to optimise 3,500 product descriptions for ecommerce in one day using ChatGPT
katarinadahlin
PRO
1
3.5k
So, you think you're a good person
axbom
PRO
2
2k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
32
2.9k
Hiding What from Whom? A Critical Review of the History of Programming languages for Music
tomoyanonymous
2
780
Measuring Dark Social's Impact On Conversion and Attribution
stephenakadiri
2
190
Transcript
Ͱ͓खܰ/-1 )JSBNBUTV!ϨτϦόηϛφʔ ը૾IUUQTHJUIVCDPNFYQMPTJPOTQB$ZCMPCNBTUFSXFCTJUFBTTFUTJNHMPHPTWH
Tsukuba, M2, NLP himkt
5-%3 w 1ZUIPOͷࣗવݴޠॲཧϥΠϒϥϦͰ͋Δ4QB$Zͷհ w Wͷ͓͠Ζػೳʹ͍ͭͯ w 4QB$ZͰຊޠςΩετΛॲཧ͢Δ
"CPVU4QB$Z ը૾IUUQTTQBDZJP https://spacy.io
IUUQTTQBDZJP "CPVU4QB$Z “Industrial-Strength NLP” w /POEFTUSVDUJWFUPLFOJ[BUJPO w /BNFEFOUJUZSFDPHOJUJPO w 4VQQPSUGPS
MBOHVBHFT w TUBUJTUJDBMNPEFMTGPSMBOHVBHFT w ʜFUD IUUQTHJUIVCDPNFYQMPTJPOTQB$Z
#BTJDVTBHFPG4QB$Z
.BOZ'FBUVSFT ը૾IUUQTTQBDZJPVTBHFGBDUTpHVSFT
'BTUFTUJOUIFXPSME w %FQFOEFODZQBSTFSͷύϑΥʔϚϯεൺֱ w จIUUQTBDMXFCPSHBOUIPMPHZ111QEG w จʹ͋Δͷ41&&%ͷදͰɼ"DDVSBDZࣗલͰ࡞ΒΕͨͷʁ w $IPJͷϕϯνϚʔΫ࣌ʹTQB$ZWະϦϦʔεͳͷͰOB
ը૾IUUQTTQBDZJPVTBHFGBDUTpHVSFT
4QFFEDPNQBSJTPOXJUIPUIFSMJCSBSJFT w ଞͷࣗવݴޠॲཧϥΠϒϥϦͱͷൺֱ w จͰͳ͘࡞ऀ͕ௐࠪͨ͠ͷ DPOEVDUFEJO w ϦϙδτϦIUUQTHJUIVCDPNFYQMPTJPOTQBDZCFODINBSLT
ը૾IUUQTTQBDZJPVTBHFGBDUTpHVSFT
.PEFMDPNQBSJTPO w ݴޠʹΑͬͯෳͷαΠζͷϞσϧ͕͋Δ FO GS FT w 104UBHHFS /&3UBHHFS
%FQFOEFODZQBSTFS w 8JUIPVUBOZQSFQSPDFTTJOH EBUBTFU ը૾IUUQTTQBDZJPVTBHFGBDUTpHVSFT
/-5,BOE4QB$Z w /-5,P⒎FSTTPNFPGUIFTBNFGVODUJPOBMJUZBTTQB$Z w *ODPNQBSJTPOUPTQB$Z /-5,UBLFTBNVDINPSF CSPBEDIVSDIBQQSPBDI w TQB$ZJTBMTPNVDINPSFQFSGPSNBODFGPDVTTFEUIBO/-5, XIFSFUIFUXPMJCSBSJFTQSPWJEFUIFTBNFGVODUJPOBMJUZ
TQB$ZTJNQMFNFOUBUJPOXJMMVTVBMMZCFGBTUFSBOENPSF BDDVSBUF Ҿ༻IUUQTTQBDZJPVTBHFGBDUTpHVSFT
0UIFSMJCSBSJFTBOETQB$Z 1ZUPSDIIUUQTHJUIVCDPNQZUPSDIQZUPSDICMPCNBTUFSEPDTTPVSDF@TUBUJDJNHQZUPSDIMPHPEBSLTWH "MMFO/-1IUUQTHJUIVCDPNBMMFOBJBMMFOOMQCMPCNBTUFSEPDTUBUJDBMMFOOMQMPHPEBSLQOH (FOTJNIUUQTHJUIVCDPN3B3F5FDIOPMPHJFTHFOTJNCMPCEFWFMPQEPDTTSDSFBENF@JNBHFTSBSFQOH $V1ZIUUQTHJUIVCDPNDVQZDVQZCMPCNBTUFSEPDTJNBHFDVQZ@MPHP@QYQOH JOUPSDIUFYU GPS(16BDDFMFSBUJPO XPSEWFDUPS
QJQFMJOF
4QB$ZWͷݸਓతʹ͖ͳػೳ w EJTQMB$ZͰ͌͢Εʔ͠ʔʁ w 4QB$ZͷՄࢹԽϞδϡʔϧ w ͖Ε͍Ͱ͍͍ײ͡ͳը૾Λ࡞ͬͯ͘ΕΔ w ݻ༗දݱநग़ͱΓड͚ղੳͷ݁ՌΛՄࢹԽͯ͘͠ΕΔ w
4QB$ZͰղੳͨ͠ΦϒδΣΫτΛͦͷ··͑Δ w 47(ܗࣜͷը૾͕ग़ྗ͞ΕΔ
8FCαΠτ্ͷ/&3ͷσϞ ը૾IUUQTFYQMPTJPOBJEFNPTEJTQMBDZFOU
4QB$ZY+VQZUFS/PUFCPPL
4QB$ZBOEຊޠ w WͰͷຊޠରԠ 13 w ຊޠܗଶૉղੳث+BOPNFΛϥοϓ͢ΔܗͰ࣮ w WͰͷܗଶૉղੳثҠߦ *TTVF
13 w ຊޠ6OJWFSTBM%FQFOEFODZσʔλ6OJ%JDͰׂ͞Ε͍ͯΔ w +BOPNFݱࡏͷͱ͜Ζ6OJ%JDʹະରԠ w .F$BCʹҠߦ
4QB$ZBOEຊޠ
4QB$ZBOEຊޠ "OTXFSVTF6OJ%JD
4QB$ZBOEຊޠ
ຊޠ/&3%FQFOEFODZQBSTJOHXJUI4QB$Z w ݁ݱࡏͰ͖ͳ͍ w 4QB$ZʹࣗͰ5BHHFS1BSTFSΛֶशͰ͖Δ ػߏ͕Έࠐ·Ε͍ͯΔ w ຊޠ6OJWFSTBM%FQFOEFODJFTެ։͞Ε͍ͯΔ IUUQTHJUIVCDPN6OJWFSTBM%FQFOEFODJFT6%@+BQBOFTF(4%ͳͲ
Ϟσϧࣗ࡞Ͱ͖ΔͷͰʂʁ
"EEJOH-BOHVBHFTVQQPSU 4QB$ZͷݴޠϞδϡʔϧͷ ίϯϙʔωϯτ ը૾IUUQTTQBDZJPVTBHFBEEJOHMBOHVBHFTTFDUJPOUSBJOJOH
"EEJOH-BOHVBHFTVQQPSU ը૾IUUQTTQBDZJPVTBHFBEEJOHMBOHVBHFTTFDUJPOUSBJOJOH 1PXFSFECZ.F$BC 4QB$ZͷֶशϞδϡʔϧ͕͑ͳ͍ʁ
ຊޠTQB$Zͷݱࡏͷʁ w ࣙॻ͕6OJ%JDͰ͋Δ͜ͱ͕ఆ͞Ε͍ͯΔ JTTVF w ʢ͓ͦΒ͘ʣଟ͘ͷڥͰ.F$BCͷσϑΥϧτࣙॻ*1"EJD w <50%0>+BQBOFTF5PLFOJ[FSͰ5BHHFSΛ࡞͍ͬͯΔ͕ɼ ͜ͷίϯετϥΫλͷҾ֎෦͔Β৮Εͳͦ͏
w ൃԻ͕ະొͷ୯ޠΛղੳ͢ΔͱΤϥʔ *1"EJD 13 w ࣙॻ͝ͱͷग़ྗͷࠩҟΛٵऩ͢ΔΠϯλʔϑΣʔε͕ඞཁʁ w /&3%FQFOEFODZ1BSTJOHͷϞσϧ·ͩଘࡏ͠ͳ͍ w ݱࡏA"MQIBUPLFOJ[BUJPOTVQQPSUA w ͔ͪॻ͖͕ඞཁͳݴޠͷରԠͲ͏͢Εʜ w தࠃޠࣅͨঢ়گͰࢭ·͍ͬͯΔ 13
͜ΕΛΕͱΓ͋͑ͣςετͰ͖ͦ͏ USBWJTͰ.F$BCΛΠϯετʔϧ͢ΔΑ͏ʹ͢Δ TQBDZNPEFMTʹ6OJ%JDΛొ͢Δ QZUIPONTQBDZEPXOMPBEKBΛඋ͢Δ /&3ͱ%FQFOEFODZ1BSTJOH͋ͱͰௐΔʜ
·ͱΊ w ࣗવݴޠॲཧϥΠϒϥϦTQB$Zͷհ w *OEVTUSZൃͷࣗવݴޠॲཧϥΠϒϥϦ w WͰೖͬͨՄࢹԽϞδϡʔϧ͕͍͍ײ͡ w EJTQMB$ZΛ͍͍ͬͨײ͡ͳՄࢹԽ w
TQB$ZͰͷຊޠςΩετॲཧ·ͩෆશ w ݱঢ়ܗଶૉղੳͷΠϯλʔϑΣʔε QJQJOTUBMMTQBDZ 3VO5IJTDPNNBOE