$30 off During Our Annual Pro Sale. View Details »
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
SappoRo.R #11「R によるThe Multilingual Eye-trackin...
Search
sakaue
February 17, 2024
Education
0
110
SappoRo.R #11「R によるThe Multilingual Eye-tracking COrpus (MECO) の探索的データ分析」
sakaue
February 17, 2024
Tweet
Share
More Decks by sakaue
See All by sakaue
RBC202003_Day2_Period5
sakaue
0
42
RBC202003_Day2_Period6
sakaue
0
100
RBC202003_Day2_Period7
sakaue
0
98
Rbootcamp202003_Day2_p8.pdf
sakaue
0
89
RBC202003_Day1_Period1
sakaue
1
79
RBC202003_Day1_Period2
sakaue
0
75
RBC202003_Day1_Period3
sakaue
0
100
RBC202003_Day1_Period4
sakaue
0
61
Other Decks in Education
See All in Education
授業レポート:共感と協調のリーダーシップ(2025年上期)
jibunal
1
180
相互コミュニケーションの難しさ
masakiokuda
0
320
1202
cbtlibrary
0
160
外国籍エンジニアの挑戦・新卒半年後、気づきと成長の物語
hypebeans
0
680
AIを使って最新研究 について調べて発表しよ う!
mickey_kubo
4
180
1125
cbtlibrary
0
140
Introdución ás redes
irocho
0
510
ThingLink
matleenalaakso
28
4.2k
Node-REDで広がるプログラミング教育の可能性
ueponx
1
220
HCI and Interaction Design - Lecture 2 - Human-Computer Interaction (1023841ANR)
signer
PRO
0
1.4k
生成AIとの付き合い方 / Generative AI and us
kaityo256
PRO
11
6.7k
AIは若者の成長機会を奪うのか?
frievea
0
140
Featured
See All Featured
Docker and Python
trallard
47
3.7k
The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs
inesmontani
PRO
2
2.8k
SEOcharity - Dark patterns in SEO and UX: How to avoid them and build a more ethical web
sarafernandez
0
89
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
25
1.7k
Scaling GitHub
holman
464
140k
Un-Boring Meetings
codingconduct
0
160
Agile Actions for Facilitating Distributed Teams - ADO2019
mkilby
0
94
Git: the NoSQL Database
bkeepers
PRO
432
66k
Build The Right Thing And Hit Your Dates
maggiecrowley
38
3k
What's in a price? How to price your products and services
michaelherold
246
13k
StorybookのUI Testing Handbookを読んだ
zakiyama
31
6.5k
What the history of the web can teach us about the future of AI
inesmontani
PRO
0
380
Transcript
ʹΑΔ ࡕ্ ୢʢ@sakaueʣ The Multilingual Eye-tracking COrpus (MECO) ͷ୳ࡧతσʔλੳ
0. ࣗݾհ
0. ࣗݾհ • ౡमಓେֶ ਓจֶ෦ ӳޠӳจֶՊ • ઐɾؔ৺ͷ͋Δ • ୈೋݴޠशಘʢSecond
Language Acquisitionʣ • ίϯϐϡʔλࢧԉޠֶֶशʢComputer Assisted Language Learningʣ • Nagoya.R ɾSappoRo.R ݩओ࠵ɼ Hiroshima.R ओ࠵ ࡕ ্ ୢ (SAKAUE, Tatsuya)
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
ӳޠؔઅͷशಘݚڀ 1. ݚڀ༰ͷհ ֶशऀ͕ΜͰ͠·͏ΞϨ... ʢղऍࠔɼൃ৴ࠔʣ
ӳޠؔઅͷशಘݚڀͷత wशಘ͞ΕΔॱং wқΛܾΊΔཁҼ 1. ݚڀ༰ͷհ
शಘ͞ΕΔॱং? • Keenan & Comrie (1977) ओ֨ʼత֨ʼؒత֨ʼલஔࢺͷ త֨ʼଐ֨ʼൺֱڃͷత֨ ͱ͍͏ॱংͰؔઅԽ͞Ε͍͢ Noun
Phrase Accessibility Hierarchy 1. ݚڀ༰ͷհ
қΛܾΊΔͷ? ྨܕ هԱ ౷ޠ Keenan & Comrie (1977) ͳͲ L1(E):
Traxler et al. (2005) L1 (E): Traxler et al. (2005) L1/L2 (J): ᖒ࡚ (2009) 1. ݚڀ༰ͷհ
ઌߦݚڀͰ༻͍ΒΕΔؔઅͷछྨ RC type examples SS The boy [that sees the
girl] chases the policeman. SO The boy [that the girl sees] chases the policeman. OS The boy chases the girl [that sees the policeman]. OO The boy chases the girl [that the policeman sees] 1. ݚڀ༰ͷհ
1. ݚڀ༰ͷհ
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
2. ࢹઢܭଌσʔλͷՄࢹԽ “The salesman that product excited was mentioned in
the newsletter.”ͱ͍͏ӳจΛ ਓͷຊਓӳޠֶशऀ͕ಡΜͩࡍͷσʔλ
ఀཹ࣌ؒɾఀཹҐஔͳͲ͕ ه͞ΕΔͷͰ͕͢... 2. ࢹઢܭଌσʔλͷՄࢹԽ
ࢹઢͷيఀཹ࣌ؒʹ جͮ͘σʔλՄࢹԽΛ Λͬͯߦ͍͍ͨ
ɹɹɹͰ͜͏ՄࢹԽ͍ͨ͠! 2. ࢹઢܭଌσʔλͷՄࢹԽ
͜͏͍͏ͷɹɹɹͰՄࢹԽ͍ͨ͠! 2. ࢹઢܭଌσʔλͷՄࢹԽ
ͪͳΈʹ ͖͞΄ͲͷՄࢹԽʹ Xेສԁͷιϑτ͕͕͕͕... ʢ͓͢͢Ίͷ package ใͳͲ͕͋Εॿ͔Γ·͢ʣ
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
Agenda 1. ݚڀ༰ͷհ 2. ࢹઢܭଌσʔλͷՄࢹԽ 3. MECO ͱ୳ࡧతσʔλੳ
3. MECO ͱ୳ࡧతσʔλੳ • MECO ͷ֓ཁ • The Multilingual Eye-tracking
COrpus • ୈ1ݴޠʢL1ʣ͕ӳޠɺୈ2ݴޠʢL2ʣ͕ӳޠͷਓ • 12छͷӳޠͷจষ (98-185ޠ)ΛಡΜͩσʔλ • OSF ʹͯσʔλެ։
3. MECO ͱ୳ࡧతσʔλੳ
3. MECO ͱ୳ࡧతσʔλੳ • MECO ͷσʔλ֓ཁ • ޠҎ֎ʹ... शख़ʢࣗݾධՁʣͳͲ •
ͬͨػࡐʢEyeLinkʣใुͷֹͳͲهࡌ • R ͷίʔυͱ .rda ܗࣜͰͷσʔλ • R Ϣʔβʔʹݶ͍ͬͯΔײ͕͍͢͝w • ඞཁͳσʔλΛͬͯ write.csv ͯ͠ OK
3. MECO ͱ୳ࡧతσʔλੳ
3. MECO ͱ୳ࡧతσʔλੳ • MECO ͷ୳ࡧσʔλੳ • "who" ΛಡΜͰ͍ͨ࣌ؒʢ6ͭͷจʹؚ·ΕΔʣ •
200ms લޙʹதԝ • ؔࢺػೳޠͱಉ͡Α͏ͳॲཧʁ • forʢػೳޠͰಉ͡จࣈʣͷͱൺֱ • ֓Ͷಉ͡ʢ͕ͩ "who" ͷ͕Ίʣ
0 200 400 600 800 du ee en fi ge
gr he it no ru sp tr lang firstrun.dur (ms) lang du ee en fi ge gr he it no ru sp tr First Run Distribution whoͷ߹
0 200 400 600 800 du ee en fi ge
gr he it no ru sp tr lang firstrun.dur (ms) lang du ee en fi ge gr he it no ru sp tr First Run Distribution for ͷ߹
3. MECO ͱ୳ࡧతσʔλੳ
3. MECO ͱ୳ࡧతσʔλੳ
3. MECO ͱ୳ࡧతσʔλੳ ※ʮϦʔτϯʯఏڙͷαʔϏεΛར༻
3. MECO ͱ୳ࡧతσʔλੳ • MECO ͷσʔλ๊͕͑Δ՝(?) • ຊਓֶशऀͷσʔλ͕ͳ͍... • ѻ͏ਓ͕গͳ͘ɺ࣮ݧڥ͕Θͳ͍͜ͱ
• MECO ͱಉ͡՝Λ͜ͳ͢ͷʹɺͦΕͳΓ ͷशख़Ͱͳ͍ͱ͍͔͠ • ಛఆͷจ๏ࣄ߲Λରʹͯ͠ੳ͢Δͷ͍͠
3. MECO ͱ୳ࡧతσʔλੳ • खʹೖΕʹ͍͘ࢹઢܭଌσʔλΛɺࣗ༝ʹɺ͔ͭɺR Ͱѻ͑ΔΑ͏ʹͯ͘͠Εͨ͜ͱ࣮ʹ͋Γ͕͍ͨ • શ͘ಉ͡ڥΛ͑ΒΕͣͱɺͦΕʹࣅͨڥ ʢWeb ΧϝϥͰࢹઢܭଌ͕Ͱ͖Δπʔϧ͋ΓʣΛ
࡞Γɺͱʹ͔͘ܭଌͯ͠ެ։ͯ͠ΈΔͷख • ࢀߟ: ʮWebGazer.js Ͱ WebΧϝϥͰ҆ՁʹΞΠτϥοΩϯάΛ͢Δʯ • https://zenn.dev/motokazu/articles/ba65d6be70370c • ੜܥ AI ͷίʔυ࡞࣮ʹ༗༻ɾ༗ೳɾେॿ͔Γ ͓ΘΓʹ
Enjoy !