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
早期うつ状態検出のためのマルチモーダル対話データセットに基づくうつ状態検出モデルの性能評価(N...
Search
Kotaro Kashihara
March 16, 2025
Research
75
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
早期うつ状態検出のためのマルチモーダル対話データセットに基づくうつ状態検出モデルの性能評価(NLP2025)
NLP2025@長崎での発表内容
Kotaro Kashihara
March 16, 2025
More Decks by Kotaro Kashihara
See All by Kotaro Kashihara
LLMアプリケーション開発 ~シンとくぽんTalkと論文要約くんの開発について~
kassy11
0
94
Other Decks in Research
See All in Research
COFFEE-Japan PROJECT Impact Report(海ノ向こうコーヒー)
ontheslope
0
2k
PGDM: Physically Guided Diffusion Model for L Downscaling
satai
2
290
Sleuthcon Keynote - How Cybercriminals (ab)use AI
fr0gger
0
190
事後確率分布の共分散について
koide3
0
150
2026年1月の生成AI領域の重要リリース&トピック解説
kajikent
0
1k
セマンティック通信勉強会 6Gに向けたデバイス間効率的な通信の技術紹介・課題・今後展望
satai
3
170
Language and AI
ayaniwa
0
140
はじまりの クエスチョンブック —余暇と豊かさにあふれた社会とは?
culturaltransition
PRO
0
520
論文紹介:HalluCitation Matters
wasyro
0
100
正規分布と最適化について
koide3
1
270
老舗ものづくり企業でリサーチが変革を起こすまで - 三菱重工DXの実践
skydats
0
200
Cross-Media Information Spaces and Architectures
signer
PRO
0
300
Featured
See All Featured
Measuring & Analyzing Core Web Vitals
bluesmoon
9
870
Chasing Engaging Ingredients in Design
codingconduct
0
220
Dominate Local Search Results - an insider guide to GBP, reviews, and Local SEO
greggifford
PRO
0
200
Introduction to Domain-Driven Design and Collaborative software design
baasie
1
860
The innovator’s Mindset - Leading Through an Era of Exponential Change - McGill University 2025
jdejongh
PRO
1
210
Building the Perfect Custom Keyboard
takai
2
800
Leveraging LLMs for student feedback in introductory data science courses - posit::conf(2025)
minecr
1
300
Heart Work Chapter 1 - Part 1
lfama
PRO
7
36k
Amusing Abliteration
ianozsvald
1
210
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
162
16k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
659
62k
How to Build an AI Search Optimization Roadmap - Criteria and Steps to Take #SEOIRL
aleyda
1
2.1k
Transcript
ദݪޭଠɹߴುढ़थɹܟଠ കݪӳ༟ɹೖᖒߤ࢙ɹத୍ཧਔɹপాपॿ ߁γϯɹ٢ాູɹদຊ ಙౡେֶେֶӃՊֶݚڀՊ࿑ಇऀ݈߁҆શػߏ ಙౡେֶҩֶ෦ಙౡେֶେֶӃࣾձ࢈ۀཧֶݚڀ෦ ૣظ͏ͭঢ়ଶݕग़ͷͨΊͷ ϚϧνϞʔμϧରσʔληοτʹجͮ͘ ͏ͭঢ়ଶݕग़ϞσϧͷੑೳධՁ $
ݚڀഎܠ • ੈքతʹετϨεͱෆ҆ʹΑΔϝϯλϧ࣬ױ͕૿Ճ͍ͯ͠Δ<> • ಛʹຊͰɺ࿑ಇऀͷϝϯλϧ࣬ױ͕૿Ճ͍ͯ͠Δ<> • ͔͠͠ɺਫ਼ਆՊҩΧϯηϥʔͷෆʹΑΓૣظൃݟ͕͍͠ 㱺ϝϯλϧෆௐͷૣظൃݟΛϚϧνϞʔμϧͳ"*ϞσϧͰସͰ͖ͳ͍͔ʁ
ݚڀత • ݈ৗऀɾϝϯλϧෆௐऀΛରͱͨ͠Χϯηϥʔͱͷ໘ஊΛ࣮ࢪ͠ɺ໘ஊதͷ ݴޠɾԻɾಈըɾ৺ഥ͔ΒͳΔϚϧνϞʔμϧσʔληοτΛߏஙɾੳ͢Δ • ߏஙͨ͠σʔληοτͱطଘσʔληοτΛΈ߹Θͤͯɺ ࠷৽ͷϚϧνϞʔμϧ͏ͭঢ়ଶݕग़ϞσϧΛֶश͠ੑೳධՁΛߦ͏ ࠷ۙ࠷ ൵͔ͬͨ͜͠ͱ
ԿͰ͔͢ʁ ࣂ͍ͬͯͨ ϖοτ͕͘ͳͬ ͨ͜ͱͰ͢ ! " σʔλऩू σʔληοτ ͏ͭঢ়ଶ ݕग़Ϟσϧ ͏ͭঢ়ଶ ݈ৗ
ؔ࿈ݚڀɿ%"*$80;<> • ͏ͭঢ়ଶݕग़ͷͨΊͷϚϧνϞʔμϧσʔληοτ %"*$80; ݴޠ ӳޠ ϞμϦςΟ ݴޠɺԻɺಈը ϥϕϧ
͏ͭঢ়ଶʹ͍ͭͯͷ Ξϯέʔτ݁Ռ ඃݧऀ ݈ৗऀ ໘ஊํ๏ 8J[BSEPG8P[๏Ͱͷ Ξόλʔͱͷ໘ஊ σʔληοτ ಛ
%"*$80;ͱͷҧ͍ 4 %"*$80; ຊݚڀ ݴޠ ӳޠ ຊޠ ϞμϦςΟ ݴޠɺԻɺಈը
ݴޠɺԻɺಈըɺ ৺ഥσʔλ ϥϕϧ ͏ͭঢ়ଶʹ͍ͭͯͷ Ξϯέʔτ݁Ռ ༷ʑͳ৺ཧঢ়ଶʢ͏ͭঢ়ଶؚΉʣ ʹ͍ͭͯͷΞϯέʔτ݁Ռ ඃݧऀ ݈ৗऀ ݈ৗऀͱ Ұ෦ϝϯλϧෆௐऀ ໘ஊํ๏ 8J[BSEPG8P[๏Ͱͷ Ξόλʔͱͷ໘ஊ 8J[BSEPG8P[๏Ͱͷ Ξόλʔͱͷ໘ஊ σʔληοτ ಛ
໘ஊํ๏ • ໘ஊ;PPNΛ௨ͯ͡ߦΘΕɺඃݧऀͷԻɾಈըΛهͨ͠ • ඃݧऀ৺ഥܭΛ͚ͯ໘ஊΛ࣮ࢪ • ΧϯηϥʔͷݟͨٴͼԻ 75VCF4UVEJPٴͼ.BHJD.JDΛ௨ͯ͡ ϦΞϧλΠϜʹมͨ͠
• ඃݧऀͱͯ͠ ʮخ͔ͬͨ͜͠ͱʯʮ൵͔ͬͨ͜͠ͱʯ ʮෲཱ͕ͬͨ͜ͱʯͷͭΛࣄલ४උ͠ɺ Χϯηϥʔ͔Βॱ࣭࣍͞Εͨ
Ξόλʔ • ໘ஊதͷΧϯηϥʔͷݟͨΛϦΞϧλΠϜͰΞόλʔʹมͨ͠ • Χϯηϥʔͷإදɾإͷ͖ͷτϥοΩϯάΛߦ͏ • ʮϋϧʯͱ͍͏໊લ͚ͩΛඃݧऀʹ໌ࣔͨ͠ • ˞໊લҎ֎ͷใʹ͍ͭͯඃݧऀʹରͯ͠໌ࣔ͠ͳ͔ͬͨ
*DPOJDCVTJOFTTNBO *O+BQBOFTF IUUQTOJ[JNBDPN*UFN%FUBJM*UFN
Ξϯέʔτ • ඃݧऀ໘ஊલɾ໘ஊޙʹΞϯέʔτʹճ • ໘ஊલɿ ඃݧऀͷ৺ཧঢ়ଶಛੑ • ໘ஊޙɿ ໘ஊதͷؾҹ
ऩू݁Ռ • ಙౡେֶٴͼಙౡେֶපӃ͔Βඃݧऀ͕ࢀՃ͠ɺݱࡏ໊ͷσʔλΛऩूࡁΈ • ͏ͪঁੑ໊ɺஉੑ໊ɺฏۉྸࡀ • 1)2ͷείΞͰ͏ͭঢ়ଶ͔൱͔Λஅͨ͠߹ɺ ݈ৗঢ়ଶ໊ɺ͏ͭঢ়ଶ໊
ಛྔͷநग़ͱ૬ؔੳ • σʔλͱΞϯέʔτ݁ՌΛ૬ؔੳ͠ɺطଘͷ৺ཧֶɾਫ਼ਆҩֶݚڀͱൺֱ • ֤ϞμϦςΟ͝ͱʹҎԼͷํ๏Ͱಛྔͷநग़Λߦͬͨ • ݴޠɿ(J/;"<>ͱຊޠධՁۃੑࣙॻ< >Λ༻͍ͯɺඃݧऀͷൃςΩετதͷ ϙδςΟϒɾωΨςΟϒͳ୯ޠٴͼׂ߹Λܭࢉ
• Իɿ0QFO4.*-&<>Λ༻͍ͯɺඃݧऀͷԻͷมԽʹؔ͢Δಛྔ ʢେ͖͞ɺϐονɺδολʔɺγϚʔʣͷฏۉɾඪ४ภࠩΛܭࢉ • ಈըɿ0QFO'BDF<>Λ༻͍ͯɺඃݧऀͷإදͷ֤ΞΫγϣϯϢχοτͷڧ͞ ͷฏۉɾඪ४ภࠩΛܭࢉ
૬ؔੳ݁Ռʢݴޠσʔλʣ • 5MCMM<>ͷ ʮ͏ͭঢ়ଶͷਓϙδςΟϒͳ ݴޠ༻͕Լ͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ • 1FOOFCBLFSΒ<>ͷ ʮ֎ަతͳਓϙδςΟϒͳ
ײޠΛଟ͘༻͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ ݴޠ ಛྔ Ξϯέʔτ ߲ ૬ؔ ϙδςΟϒͳ ༻ݴͷׂ߹ ("% ಉ্ 1)2 ಉ্ 4$"4 ϙδςΟϒͳ ༻ݴͷ #*( ֎ੑ ϙδςΟϒͳ ୯ޠͷ ಉ্
૬ؔੳ݁Ռʢݴޠσʔλʣ • 5MCMM<>ͷ ʮ͏ͭঢ়ଶͷਓϙδςΟϒͳ ݴޠ༻͕Լ͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ • 1FOOFCBLFSΒ<>ͷ ʮ֎ަతͳਓϙδςΟϒͳ
ײޠΛଟ͘༻͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ ݴޠ ಛྔ Ξϯέʔτ ߲ ૬ؔ ϙδςΟϒͳ ༻ݴͷׂ߹ ("% ಉ্ 1)2 ಉ্ 4$"4 ϙδςΟϒͳ ༻ݴͷ #*( ֎ੑ ϙδςΟϒͳ ୯ޠͷ ಉ্
૬ؔੳ݁Ռʢݴޠσʔλʣ • 5MCMM<>ͷ ʮ͏ͭঢ়ଶͷਓϙδςΟϒͳ ݴޠ༻͕Լ͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ • 1FOOFCBLFSΒ<>ͷ ʮ֎తͳਓϙδςΟϒͳ
ײޠΛଟ͘༻͢Δʯ ͱ͍͏݁Ռͱಉ༷ͷ݁Ռ ݴޠ ಛྔ Ξϯέʔτ ߲ ૬ؔ ϙδςΟϒͳ ༻ݴͷׂ߹ ("% ಉ্ 1)2 ಉ্ 4$"4 ϙδςΟϒͳ ༻ݴͷ #*( ֎ੑ ϙδςΟϒͳ ୯ޠͷ ಉ্
૬ؔੳ݁ՌʢԻσʔλʣ • (BMJMMJΒ<>ͷ ʮࣾަෆ҆Λ࣋ͭਓɺ جຊपʢϐονʣ͕େ͖͘ɺ ͷେ͖͕͞খ͘͞ͳΔʹ͋Δʯ ͱ͍͏݁ՌͱҰ෦Ұக • ,PUPWΒ<>ͷ
ʮ͏ͭঢ়ଶͷਓ ͷมԽΤωϧΪʔ͕͘͠ͳΔʯ ͱ͍͏݁ՌͱҰ෦Ұக Ի ಛྔ Ξϯέʔτ ߲ ૬ؔ ͷϐονͷ ฏۉ -4"4 ͷେ͖͞ͷ ඪ४ภࠩ 1)2 ͷϐονͷ ඪ४ภࠩ ಉ্
૬ؔੳ݁ՌʢԻσʔλʣ • (BMJMMJΒ<>ͷ ʮࣾަෆ҆Λ࣋ͭਓɺ جຊपʢϐονʣ͕େ͖͘ɺ ͷେ͖͕͞খ͘͞ͳΔʹ͋Δʯ ͱ͍͏݁ՌͱҰ෦Ұக • ,PUPWΒ<>ͷ
ʮ͏ͭঢ়ଶͷਓ ͷมԽΤωϧΪʔ͕͘͠ͳΔʯ ͱ͍͏݁ՌͱҰ෦Ұக Ի ಛྔ Ξϯέʔτ ߲ ૬ؔ ͷϐονͷ ฏۉ -4"4 ͷେ͖͞ͷ ඪ४ภࠩ 1)2 ͷϐονͷ ඪ४ภࠩ ಉ্
૬ؔੳ݁ՌʢԻσʔλʣ • (BMJMMJΒ<>ͷ ʮࣾަෆ҆Λ࣋ͭਓɺ جຊपʢϐονʣ͕େ͖͘ɺ ͷେ͖͕͞খ͘͞ͳΔʹ͋Δʯ ͱ͍͏݁ՌͱҰ෦Ұக • ,PUPWΒ<>ͷ
ʮ͏ͭঢ়ଶͷਓ ͷมԽΤωϧΪʔ͕͘͠ͳΔʯ ͱ͍͏݁ՌͱҰ෦Ұக Ի ಛྔ Ξϯέʔτ ߲ ૬ؔ ͷϐονͷ ฏۉ -4"4 ͷେ͖͞ͷ ඪ४ภࠩ 1)2 ͷϐονͷ ඪ४ภࠩ ಉ্
૬ؔੳ݁Ռʢಈըσʔλʣ • #*(։์ੑͷߴ͍ඃݧऀɺ ൺֱతϦϥοΫεͯ͠໘ஊΛ ड͚͍ͯͨ͜ͱΛࣔࠦ • #*(ۈษੑͷείΞͱ ΄ͱΜͲͷΞΫγϣϯϢχοτͷڧ͞ͷ ฏۉɾඪ४ภࠩͱʹෛͷ૬͕ؔݟΒΕͨ
㱺#*(ۈษੑͷߴ͍ඃݧऀ ൺֱతແදͰ໘ஊΛड͚͍ͯͨ͜ͱΛࣔࠦ ಈը ಛྔ Ξϯέʔτ ߲ ૬ؔ ʮඑΛԼ͛Δʯ ڧ͞ͷฏۉ #*( ։์ੑ ʮᛐΛۓுͤ͞ Δʯ ڧ͞ͷฏۉ ಉ্ ʮඑͷ֎ଆΛ্ ͛Δʯ ڧ͞ͷฏۉ ಉ্
࣮ݧɿલॲཧͱಛྔநग़ • ߏஙͨ͠σʔληοτͱ%"*$80;ΛΈ߹ΘֶͤͨशΛ࣮ࢪ 㱺ͭͷσʔληοτͰ౷Ұͨ͠લॲཧɾಛྔநग़Λ࣮ࢪ • લॲཧɿੜσʔλ͔Βඃݧऀͷൃ۠ؒͷԻɾಈըϑϨʔϜͷΈΛநग़ • ಛྔநग़ɿ •
ԻಛྔɿࣄલֶशࡁΈϞσϧ7((JTI<>ͷ࠷ऴϕΫτϧ • ಈըಛྔɿ0QFO'BDF͔ΒಘΒΕΔإϥϯυϚʔΫ
࣮ݧɿϞσϧͱֶशϥϕϧ • 1)2ٴͼ1)2ͷ݁ՌΛͱʹɺϥϕϧʢ͏ͭঢ়ଶɾ݈ৗঢ়ଶʣΛ࡞ • ֶशɾݕূɾςετηοτΛʹׂ • ࠷৽ͷϚϧνϞʔμϧ͏ͭঢ়ଶݕग़ϞσϧͰ͋Δ%FQ.BNCB<>Λར༻ • %"*$80;Ͱࣄલֶश㱺ߏஙͨ͠σʔληοτͰϑΝΠϯνϡʔχϯά
• ߏஙͨ͠σʔληοτ୯ମͰͷֶशͱൺֱ %"*$80; ߏஙͨ͠σʔληοτ σʔληοτ ࣄલֶश ϑΝΠϯνϡʔχϯά ! " ͏ͭঢ়ଶ ݈ৗ ࣄલֶशϞσϧ ϑΝΠϯνϡʔχϯά Ϟσϧ
࣮ݧ݁Ռ • %"*$80;ͷࣄલֶशΛߦ͏͜ͱͰɺ΄ͱΜͲͷධՁࢦඪ͕վળ • ϥϕϧͷภΓ͕͋ΔͨΊɺ୯ମͰͷֶश࣌ʹ࠶ݱ͕͔ͬͨ • ࣄલֶशΛߦ͏͜ͱͰେ෯ʹվળ͞Εʮ͏ͭঢ়ଶΛݟಀ͞ͳ͍Ϟσϧʯʹ
࣮ݧ݁Ռ • ҰํಛҟԼ͠ɺޡͬͯ͏ͭঢ়ଶΛݕग़ͯ͠͠·͏͕૿͑ͨ • 㱺ࠓޙσʔλΛ֦ॆͯ͠Ͱ͖Δ͚ͩϥϕϧͷภΓΛͳ͘͠ɺ ࠶ݱɾಛҟͱʹߴ͍ϞσϧͷߏஙΛࢦ͢
ࠓޙͷ՝ σʔλ֦ॆ wܧଓͯ͠σʔλΛऩू w͏ͭঢ়ଶɾ݈ৗঢ়ଶͷϥϕϧͷภΓΛͰ͖Δ͚ͩͳ͘͢ ٬؍Ξϊςʔγϣϯͷ༩ wओ؍ϥϕϧʢඃݧऀͷΞϯέʔτ݁Ռʣ͔͍͍ͯ͠ͳ͍ wୈࡾऀɾઐՈʢਫ਼ਆՊҩͳͲʣʹΑΔΞϊςʔγϣϯΛ࣮ࢪ
Ϟσϧֶशͷ w1)2Ҏ֎ͷΞϯέʔτ݁Ռ٬؍Ξϊςʔγϣϯར༻ֶͯ͠श
·ͱΊ • ͏ͭঢ়ଶݕग़ͷͨΊͷຊޠͷϚϧνϞʔμϧσʔληοτΛߏஙɾੳ • ߏஙͨ͠σʔληοτͱطଘσʔληοτΛΈ߹ΘͤͯϞσϧΛֶश • ࠓޙɺσʔλ֦ॆσʔλʹର͢Δ٬؍ΞϊςʔγϣϯΛ࣮ࢪ͢Δ
ࢀߟจݙ <>8PSME)FBMUI0SHBOJ[BUJPO$PWJEQBOEFNJDUSJHHFSTJODSFBTFJOQSFWBMFODFPGBOYJFUZBOEEFQSFTTJPOXPSMEXJEF IUUQTXXXXIPJOUOFXTJUFNDPWJEQBOEFNJDUSJHHFSTJODSFBTFJOQSFWBMFODFPGBOYJFUZBOEEFQSFTTJPO XPSMEXJEF "DDFTTFE+BOVBSZ <>+PC૯ݚʹΑΔʰϝϯλϧέΞͷҙࣝௐࠪʱΛ࣮ࢪզຫͦ͜ඒಙ ऑԻు࿐zܴzׂ͕ෆௐzݴ͑ͳ͍z IUUQTQSUJNFTKQ NBJOIUNMSEQIUNM"DDFTTFE+VOF
<>(SBUDI+ "SUTUFJO3 -VDBT(. 4USBUPV( 4DIFSFS4 /B[BSJBO" 8PPE3 #PCFSH+ %F7BVMU% .BSTFMMB4 5SBVN%35IF %JTUSFTT"OBMZTJT*OUFSWJFX$PSQVTPG)VNBOBOE$PNQVUFS*OUFSWJFXT*O1SPDFFEJOHTPG-3&$.BZ QQ <>দా(J/;"6OJWFSTBM%FQFOEFODJFTʹΑΔ࣮༻తຊޠղੳࣗવݴޠॲཧ7PMVNF/VNCFS <>খྛͷͧΈɼס݈ଠɼদຊ༟࣏ɼཱੴ݈ೋɼౡढ़Ұҙݟநग़ͷͨΊͷධՁදݱͷऩूࣗવݴޠॲཧɼ7PM /P QQ <>౦ࢁণ ס݈ଠ দຊ༟࣏ ड़ޠͷબબੑʹண໊ͨ͠ࢺධՁۃੑͷ֫ಘ ݴޠॲཧֶձୈճ࣍େձจू QQ <>MPSJBO&ZCFO .BSUJO8ÖMMNFS #KÖSO4DIVMMFSPQFO4.*-&5IF.VOJDI7FSTBUJMFBOE'BTU0QFO4PVSDF"VEJP'FBUVSF &YUSBDUPS 1SPD"$..VMUJNFEJB .. "$. 'MPSFODF *UBMZ *4#/ QQ <>0QFO'BDF'BDJBM#FIBWJPS"OBMZTJT5PPMLJU5BEBT#BMUSVãBJUJT "NJS;BEFI :BP$IPOH-JN BOE-PVJT1IJMJQQF.PSFODZ *&&&*OUFSOBUJPOBM$POGFSFODFPO"VUPNBUJD'BDFBOE(FTUVSF3FDPHOJUJPO
ࢀߟจݙ <>,BUSJOF#OOFMBOE5MCMM-JOHVJTUJDGFBUVSFTJOEFQSFTTJPOBNFUBBOBMZTJT+PVSOBMPG-BOHVBHF8PSLT4QSPHWJEFOTLBCFMJHU 4UVEFOUFSUJETTLSJGU 7PM /P Qr %FD <>1FOOFCBLFS +8
.FIM .3 /JFEFSIP⒎FS , 1TZDIPMPHJDBMBTQFDUTPGOBUVSBMMBOHVBHFVTFPVSXPSET PVSTFMWFT "OOVBMSFWJFXPGQTZDIPMPHZ <>(BMJMJ - "NJS 0 (JMCPB4DIFDIUNBO & "DPVTUJDQSPQFSUJFTPGEPNJOBODFBOESFRVFTUVUUFSBODFTJOTPDJBMBOYJFUZ +PVSOBMPG4PDJBMBOE$MJOJDBM1TZDIPMPHZ r<>3PNBO,PUPW 8BLJ[B(BNF[ 'SBOL4DINJEU BOE%BWJE8BUTPO-JOLJOH zCJHzQFSTPOBMJUZUSBJUTUPBOYJFUZ EFQSFTTJWF BOETVCTUBODFVTFEJTPSEFST"NFUBBOBMZTJT1TZDIPMPHJDBMCVMMFUJO 7PM QQr <>4IBXO)FSTIFZ 4PVSJTI$IBVEIVSJ %BOJFM18&MMJT +PSU'(FNNFLF "SFO+BOTFO 3$IBOOJOH.PPSF .BOPK1MBLBM %FWJO1MBUU 3JG"4BVSPVT #SZBO4FZCPME .BMDPMN4MBOFZ 3PO+8FJTT BOE,FWJO88JMTPO$OOBSDIJUFDUVSFTGPSMBSHFTDBMFBVEJP DMBTTJpDBUJPO*&&&*OUFSOBUJPOBM$POGFSFODFPO"DPVTUJDT 4QFFDIBOE4JHOBM1SPDFTTJOH *$"441 QQr <>+JBYJO:F +VOQJOH;IBOH BOE)POHNJOH4IBO%FQNBNCB1SPHSFTTJWFGVTJPONBNCBGPSNVMUJNPEBMEFQSFTTJPOEFUFDUJPO*O *$"441*&&&*OUFSOBUJPOBM$POGFSFODFPO"DPVTUJDT 4QFFDIBOE4JHOBM1SPDFTTJOH *$"441 )ZEFSBCBE *OEJB "QSJM QQr <>$IFO 3 $IFO 9 /J # (F : 4JN4XBQ"O&⒏DJFOU'SBNFXPSL'PS)JHI'JEFMJUZ'BDF4XBQQJOH1SPDFFEJOHTPGUIF UI"$.*OUFSOBUJPOBM$POGFSFODFPO.VMUJNFEJB