DEIM Forum 2017 G7-1 Songrium N N Songrium N N N Web 1. Web Web UG

Size: px

Start display at page:

Download "DEIM Forum 2017 G7-1 Songrium N N Songrium N N N Web 1. Web Web UG"

ゆりかいなおか
4 years ago
Views:

1 DEIM Forum 2017 G7-1 Songrium N {k.tsukuda,ksuke-ishida,masahiro.hamasaki,m.goto}@aist.go.jp N Songrium N N N Web 1. Web Web UGC User Generated Content YouTube 1 2 [1] [2] [3] 3D 3D web Thigiverse 3 [4] N N+1 N [5] N Web [6] Web [7] N N [6] VOCALOID [8] Web Songrium

2 4 N N 2. [N ] Eto [9] 3D Web 10.4% Cheliotic [10] ccmixter 5 N N Hamasaki [1] Web [] [11] [12] [13] Saito [14] Iwata [15] Shared Cascade Poisson Process SCPP Tanaka [16] SCPP SCPP SCPP Tanaka [16] N SCPP 3. [15], [16] N [1], [9], [10] rich-get-richer [17] [1] N [6] [6] D T I Web 0 T (t p ij, up ij ) i I j u p ij U tp ij i U j < j t p ij < = tp ij J i i i D i = {(t p ij, up ij )}J i j=1 D = {D i} i I Web (t o ik, r o ik) i I k i t o ik r o ik K o i i i O i = {(t o ik, r o ik)} Ko i k=1 O = {O i} i I

1 (a) (b) (c) (d) (e) u t i (t c ik, r c ik) i k K c i i i C i = {(t c ik, r c ik)} Kc i k=1 C i i (t c ik, rik) c (t c ik, rc ik ) C = {C i } i I 3. 2

3 1 (a) (b) (c) (d) (e) u t i (t c ik, r c ik) i k K c i i i C i = {(t c ik, r c ik)} Kc i k=1 C i i (t c ik, rik) c (t c ik, rc ik ) C = {C i } i I i u i u i i i 0 T 1 a t t i i u i f i (u) = α i θ 0u. (1) α i > = 0 i θ 0u > = 0 u θ0u = 1 1 a u U α iθ 0u i u i 3. 1 i u t i rb(r )ω iθ 1ue γo(t t ) if t < t h o(i,t,r )(t, u) = 0 otherwise. r i t rb Web Web [7] rb(r ) = 1 ω r i > = 0 i θ 1u > = 0 u u U θ 1u = 1 e γ o(t t ) γ o > = 0 (2) 1 b i r t o i1 t o i1 rb(r )ω i θ 1u h u i u i i t r u t i rb(r )σ i θ 2u e γ d(t t ) if t < t h d(i,t,r )(t, u) = 0 otherwise. σ i > = 0 i θ 2u > = 0 u u U θ 2u = 1 e γ d(t t ) γ d > = 0 1 c d i (3)

4 1 c r t c i1 rb(r )σ iθ 2u h u t i λ i(t, u) = f i(u) + (t,r ) O it h o(i,t,r )(t, u) + (t,r ) C it h d(i,t,r )(t, u). (4) O it = {(t, r ) (t, r ) O i and t < t} C it = {(t, r ) (t, r ) C i and t < t} λ i (t, u) 1 e h 3 4. D O C D P (D O, C, α, ω, σ, Θ, γ) = ( T exp λ i (t, u)dt i I 0 u U ) Ji j=1 λ i (t p ij, tp ij ). (5) α = {α i } i I ω = {ω i } i I σ = {σ i } i I Θ = {θ u } u U + θ u = {θ u u} u U γ = {γ o, γ d } U + = {0, 1, 2} T 0 u U λ i(t, u)dt = α it + ωi γ o + σ i γ d K c i k=1 K o i k=1 rb(rik) (1 o e γ o(t t o )) ik ( rb(rik) c 1 e γ d(t t c )) ik. (6) Iwata [15] i j z ij z ij {0, 1,, O it + C it } = 0 1 < = z ij < = O it O it + 1 < = z ij < = O it + C it 4 λ i (t, u) = z λ i(t, u, z) λ i (t, u, z) = (7) f i (u) if z = 0 h o(i,t o iz,r iz o (t, u) if 1 < = z < = O it h d(i,t c iz,rc iz ) (t, u) if O it\u + 1 < = z < = O it\u + C it\u. 1 I O D C U 5, , , ,086 31, ,828 31,677 76,457 3, ,024 9,316 52,514 1,066 z = z O it (5) (6) (7) D Z = {{z ij } J i j=1 } i I P (D, Z O, C, α, ω, σ, Θ, γ) (8) = exp α i T + ω K i o i rb(rik) (1 o e γ o(t t o )) ik γ o i I k=1 + σ K i c i rb(rik) (1 c e γ d(t t c )) J i ik λ i (t p ij γ, tp ij, z ij). d k=1 j=1 α i α a b P (α i a, b) α a 1 i exp( bα i) Iwata [15] a = b = 1 ω i ω σ i σ θ u Θ β P (θ u β) u U θβ 1 u u α ω σ Θ γ P (D, Z O, C, γ, β, a, b) = P (D, Z O, C, α, ω, σ, Θ, γ)p (α a, b) P (ω a, b)p (σ a, b)p (Θ β)dαdωdσdθ. (9) (9) EM z ij Z E z ij z ij P (z ij = y D, Z \ij, O, C, γ, β, a, b) P ( D, Z \ij, z ij = y O, C, γ, β, a, b ) P ( D \ij, Z \ij O, C, γ, β, a, b ). (10) y {0, 1,, O it + C it } \ij i j M γ Z 5. Songrium Songrium 5. 1 Songrium N

$2 3 Algorithm 1 i j E con E pop Require: P (z ij D, Z \ij, O, C, γ, β, a, b) 1: E con P (z ij = 0$ $= y D, Z \ij, O, C, γ, β, a, b) 5: y y + 1 6: end while 7: return E con, E pop N 2017 2 15 62$

5 2 3 Algorithm 1 i j E con E pop Require: P (z ij D, Z \ij, O, C, γ, β, a, b) 1: E con P (z ij = 0 D, Z \ij, O, C, γ, β, a, b), E pop 0 2: y 1 3: while y < = O it + C it do 4: E pop E pop + P (z ij = y D, Z \ij, O, C, γ, β, a, b) 5: y y + 1 6: end while 7: return E con, E pop N VOCALOID [8] Songrium Songrium 5. 2 Songrium 2 3 i

5. 2. 1 (10) Algorithm 1 i j 3. 2. 1 E con 3. 2. 2 3. 2. 3 E pop i E con E pop i SUME con SUME pop 1 1 i V (i) ( ) SUME pop V (i) = 2 0.

2 u 4 2 3 4 5 Algorithm 1 E con E pop CE con CE pop u i CE con CE pop i i CE con CE pop CSUME con CSUME pop 1

6 (10) Algorithm 1 i j E con E pop i E con E pop i SUME con SUME pop 1 1 i V (i) ( ) SUME pop V (i) = (11) SUME pop + SUME con u Algorithm 1 E con E pop CE con CE pop u i CE con CE pop i i CE con CE pop CSUME con CSUME pop 1 1 i H(i) ( ) CSUME pop H(i) = (12) CSUME pop + CSUME con [1] N 6 i j

7 6 (10) z ij y y = 0 1 < = y < = O it j

2. 1 SUME con SUME pop SUME pop SUME con

2 3 4 8 8 u i j 3 Algorithm 2 E f E ho E

7 7 6 (10) z ij y y = 0 1 < = y < = O it j 1 y j j < j j j j j N i SUME con SUME pop SUME pop SUME con +SUME pop u i j 3 Algorithm 2 E f E ho E hd u E f E ho E hd SUME f SUME ho SUME hd 1 u

$Algorithm 2 i j Require: P (z ij D, Z \ij, O, C, γ, β, a, b) 1: E f P (z ij = 0 D, Z \ij, O, C, γ, β, a, b), E ho 0, E hd 0 2: y 1 3: while y < = O it + C it do 4: if y < = O it then 5: E ho E ho + P$

8 Algorithm 2 i j Require: P (z ij D, Z \ij, O, C, γ, β, a, b) 1: E f P (z ij = 0 D, Z \ij, O, C, γ, β, a, b), E ho 0, E hd 0 2: y 1 3: while y < = O it + C it do 4: if y < = O it then 5: E ho E ho + P (z ij = y D, Z \ij, O, C, γ, β, a, b) 6: else 7: E hd E hd + P (z ij = y D, Z \ij, O, C, γ, β, a, b) 8: end if 9: y y : end while 11: return E f, E ho, E hd N Web Songrium Web 15H06887 JST CREST ACCEL [1] M. Hamasaki, H. Takeda and T. Nishimura: Network analysis of massively collaborative creation of multimedia contents: Case study of hatsune miku videos on nico nico douga, UXTV, pp (2008). [2] C. Cayari: The youtube effect: How youtube has provided new ways to consume, create, and share music, International Journal of Education & the Arts, 12, 6, pp (2011). [3] L. A. Liikkanen and A. Salovaara: Music on youtube: user engagement with traditional, user-appropriated and derivative videos, Computers in Human Behavior, 50, pp (2015). [4] S. Papadimitriou and E. E. Papalexakis: Towards laws of the 3d-printable design web, WebSci, pp (2014). [5] M. Goto: Grand challenges in music information research, Dagstuhl Follow-Ups: Multimodal Music Processing, 3, pp (2012). [6] K. Tsukuda, M. Hamasaki and M. Goto: Why did you cover that song?: Modeling n-th order derivative creation with content popularity, CIKM, pp (2016). [7] T. Joachims, L. Granka, B. Pan, H. Hembrooke and G. Gay: Accurately interpreting clickthrough data as implicit feedback, SIGIR, pp (2005). [8] H. Kenmochi and H. Ohshita: Vocaloid - commercial singing synthesizer based on sample concatenation, IN- TERSPEECH, pp (2007). [9] K. Eto, M. Hamasaki, K. Watanabe, Y. Kawasaki and T. Nishimura: Modulobe: A creation and sharing platform for articulated models with complex motion, ACE, pp (2008). [10] G. Cheliotis and J. Yew: An analysis of the social structure of remix culture, C&T, pp (2009). [11] X. Song, Y. Chi, K. Hino and B. L. Tseng: Information flow modeling based on diffusion rate for prediction and ranking, WWW, pp (2007). [12] X. Song, B. L. Tseng, C.-Y. Lin and M.-T. Sun: Personalized recommendation driven by information flow, SIGIR, pp (2006). [13] J. Yang and J. Leskovec: Modeling information diffusion in implicit networks, ICDM, pp (2010). [14] K. Saito, M. Kimura, K. Ohara and H. Motoda: Learning continuous-time information diffusion model for social behavioral data analysis, ACML, pp (2009). [15] T. Iwata, A. Shah and Z. Ghahramani: Discovering latent influence in online social activities via shared cascade poisson processes, KDD, pp (2013). [16] Y. Tanaka, T. Kurashima, Y. Fujiwara, T. Iwata and H. Sawada: Inferring latent triggers of purchases with consideration of social effects and media advertisements, WSDM, pp (2016). [17] A.-L. Barabási and R. Albert: Emergence of scaling in random networks, Science, 286, 5439, pp (1999).

Wikipedia YahooQA MAD 4)5) MAD Web 6) 3. YAMAHA 7) 8) 2 3 4 5 6 2. Vocaloid2 2006 1 PV 2009 1 1100 200 YouTube 1 minato minato ussy 3D MAD F EDis ussy

Wikipedia YahooQA MAD 4)5) MAD Web 6) 3. YAMAHA 7) 8) 2 3 4 5 6 2. Vocaloid2 2006 1 PV 2009 1 1100 200 YouTube 1 minato minato ussy 3D MAD F EDis ussy 1, 2 3 1, 2 Web Fischer Social Creativity 1) Social Creativity CG Network Analysis of an Emergent Massively Collaborative Creation Community Masahiro Hamasaki, 1, 2 Hideaki Takeda 3 and Takuichi Nishimura