IPSJ SIG Technical Report Vol.2017-CSEC-77 No.4 Vol.2017-IOT-37 No /5/25 1,a) 1,b) 1,2,c) 1,d) BtoC EC EC EC EC ImageNet Inception-v3 EC A searc

Similar documents
SummaryJP.pdf

untitled

DEIM Forum 2019 J

Convolutional Neural Network A Graduation Thesis of College of Engineering, Chubu University Investigation of feature extraction by Convolution

untitled

(a) 1 (b) 3. Gilbert Pernicka[2] Treibitz Schechner[3] Narasimhan [4] Kim [5] Nayar [6] [7][8][9] 2. X X X [10] [11] L L t L s L = L t + L s

DEIM Forum 2018 C5-2 CGM Youtube pixiv comico CGM Co

IPSJ SIG Technical Report Vol.2015-MPS-103 No.29 Vol.2015-BIO-42 No /6/24 Deep Convolutional Neural Network 1,a) 1,b),c) X CT (Computer Aided D

IPSJ SIG Technical Report Vol.2009-CVIM-167 No /6/10 Real AdaBoost HOG 1 1 1, 2 1 Real AdaBoost HOG HOG Real AdaBoost HOG A Method for Reducing

Haiku Generation Based on Motif Images Using Deep Learning Koki Yoneda 1 Soichiro Yokoyama 2 Tomohisa Yamashita 2 Hidenori Kawamura Scho

& Vol.5 No (Oct. 2015) TV 1,2,a) , Augmented TV TV AR Augmented Reality 3DCG TV Estimation of TV Screen Position and Ro

IPSJ SIG Technical Report Vol.2010-CVIM-170 No /1/ Visual Recognition of Wire Harnesses for Automated Wiring Masaki Yoneda, 1 Ta

1 Web [2] Web [3] [4] [5], [6] [7] [8] S.W. [9] 3. MeetingShelf Web MeetingShelf MeetingShelf (1) (2) (3) (4) (5) Web MeetingShelf

IPSJ SIG Technical Report Vol.2013-CVIM-187 No /5/30 1,a) 1,b), 1,,,,,,, (DNN),,,, 2 (CNN),, 1.,,,,,,,,,,,,,,,,,, [1], [6], [7], [12], [13]., [

IPSJ SIG Technical Report Vol.2017-CVIM-207 No /5/10 GAN 1,a) 2,b) Generative Adversarial Networks GAN GAN CIFAR-10 10% GAN GAN Stacked GAN Sta

IPSJ SIG Technical Report Vol.2012-CG-149 No.13 Vol.2012-CVIM-184 No /12/4 3 1,a) ( ) DB 3D DB 2D,,,, PnP(Perspective n-point), Ransa

[2] OCR [3], [4] [5] [6] [4], [7] [8], [9] 1 [10] Fig. 1 Current arrangement and size of ruby. 2 Fig. 2 Typography combined with printing

28 TCG SURF Card recognition using SURF in TCG play video

TF-IDF TDF-IDF TDF-IDF Extracting Impression of Sightseeing Spots from Blogs for Supporting Selection of Spots to Visit in Travel Sat

Microsoft Word - toyoshima-deim2011.doc

THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE. TRECVID2012 Instance Search {sak

17 Proposal of an Algorithm of Image Extraction and Research on Improvement of a Man-machine Interface of Food Intake Measuring System

SICE東北支部研究集会資料(2017年)

2 Fig D human model. 1 Fig. 1 The flow of proposed method )9)10) 2.2 3)4)7) 5)11)12)13)14) TOF 1 3 TOF 3 2 c 2011 Information

1 Fig. 1 Extraction of motion,.,,, 4,,, 3., 1, 2. 2.,. CHLAC,. 2.1,. (256 ).,., CHLAC. CHLAC, HLAC. 2.3 (HLAC ) r,.,. HLAC. N. 2 HLAC Fig. 2

35_3_9.dvi

_314I01BM浅谷2.indd

03_特集2_3校_0929.indd

3 2 2 (1) (2) (3) (4) 4 4 AdaBoost 2. [11] Onishi&Yoda [8] Iwashita&Stoica [5] 4 [3] 3. 3 (1) (2) (3)

Duplicate Near Duplicate Intact Partial Copy Original Image Near Partial Copy Near Partial Copy with a background (a) (b) 2 1 [6] SIFT SIFT SIF

IPSJ SIG Technical Report Secret Tap Secret Tap Secret Flick 1 An Examination of Icon-based User Authentication Method Using Flick Input for

Vol.55 No (Jan. 2014) saccess 6 saccess 7 saccess 2. [3] p.33 * B (A) (B) (C) (D) (E) (F) *1 [3], [4] Web PDF a m

Silhouette on Image Object Silhouette on Images Object 1 Fig. 1 Visual cone Fig. 2 2 Volume intersection method Fig. 3 3 Background subtraction Fig. 4

IPSJ SIG Technical Report iphone iphone,,., OpenGl ES 2.0 GLSL(OpenGL Shading Language), iphone GPGPU(General-Purpose Computing on Graphics Proc

Vol. 42 No. SIG 8(TOD 10) July HTML 100 Development of Authoring and Delivery System for Synchronized Contents and Experiment on High Spe

TA3-4 31st Fuzzy System Symposium (Chofu, September 2-4, 2015) Interactive Recommendation System LeonardoKen Orihara, 1 Tomonori Hashiyama, 1

3_23.dvi

IPSJ SIG Technical Report Vol.2010-GN-74 No /1/ , 3 Disaster Training Supporting System Based on Electronic Triage HIROAKI KOJIMA, 1 KU

1: 2: 3: 4: 2. 1 Exploratory Search [4] Exploratory Search 2. 1 [7] [8] [9] [10] Exploratory Search

IPSJ SIG Technical Report Vol.2014-GN-90 No.16 Vol.2014-CDS-9 No.16 Vol.2014-DCC-6 No /1/24 1,a) 2,b) 2,c) 1,d) QUMARION QUMARION Kinect Kinect

,4) 1 P% P%P=2.5 5%!%! (1) = (2) l l Figure 1 A compilation flow of the proposing sampling based architecture simulation

IPSJ SIG Technical Report Vol.2011-EC-19 No /3/ ,.,., Peg-Scope Viewer,,.,,,,. Utilization of Watching Logs for Support of Multi-

IPSJ SIG Technical Report Vol.2009-DPS-141 No.20 Vol.2009-GN-73 No.20 Vol.2009-EIP-46 No /11/27 1. MIERUKEN 1 2 MIERUKEN MIERUKEN MIERUKEN: Spe

Fig. 3 3 Types considered when detecting pattern violations 9)12) 8)9) 2 5 methodx close C Java C Java 3 Java 1 JDT Core 7) ) S P S

1., 1 COOKPAD 2, Web.,,,,,,.,, [1]., 5.,, [2].,,.,.,, 5, [3].,,,.,, [4], 33,.,,.,,.. 2.,, 3.., 4., 5., ,. 1.,,., 2.,. 1,,

24 Region-Based Image Retrieval using Fuzzy Clustering

IPSJ-TOM

知識ベースCFD

Vol.53 No (Mar. 2012) 1, 1,a) 1, 2 1 1, , Musical Interaction System Based on Stage Metaphor Seiko Myojin 1, 1,a

TCP/IP IEEE Bluetooth LAN TCP TCP BEC FEC M T M R M T 2. 2 [5] AODV [4]DSR [3] 1 MS 100m 5 /100m 2 MD 2 c 2009 Information Processing Society of

4. C i k = 2 k-means C 1 i, C 2 i 5. C i x i p [ f(θ i ; x) = (2π) p 2 Vi 1 2 exp (x µ ] i) t V 1 i (x µ i ) 2 BIC BIC = 2 log L( ˆθ i ; x i C i ) + q

: ( 1) () 1. ( 1) 2. ( 1) 3. ( 2)

1 Table 1: Identification by color of voxel Voxel Mode of expression Nothing Other 1 Orange 2 Blue 3 Yellow 4 SSL Humanoid SSL-Vision 3 3 [, 21] 8 325

( ) [1] [4] ( ) 2. [5] [6] Piano Tutor[7] [1], [2], [8], [9] Radiobaton[10] Two Finger Piano[11] Coloring-in Piano[12] ism[13] MIDI MIDI 1 Fig. 1 Syst

IPSJ SIG Technical Report Vol.2019-IOT-44 No /3/8 CRIU HTTP 1,a) 2 PC SNS Web FastContainer HTTP HTTP Web HTTP Web Web Low-cost and High Perfor

IPSJ SIG Technical Report Vol.2012-CG-148 No /8/29 3DCG 1,a) On rigid body animation taking into account the 3D computer graphics came

1 1 CodeDrummer CodeMusician CodeDrummer Fig. 1 Overview of proposal system c

B HNS 7)8) HNS ( ( ) 7)8) (SOA) HNS HNS 4) HNS ( ) ( ) 1 TV power, channel, volume power true( ON) false( OFF) boolean channel volume int

P2P P2P peer peer P2P peer P2P peer P2P i

IHI Robust Path Planning against Position Error for UGVs in Rough Terrain Yuki DOI, Yonghoon JI, Yusuke TAMURA(University of Tokyo), Yuki IKEDA, Atsus

2). 3) 4) 1.2 NICTNICT DCRA Dihedral Corner Reflector micro-arraysdcra DCRA DCRA DCRA 3D DCRA PC USB PC PC ON / OFF Velleman K8055 K8055 K8055

Vol.54 No (July 2013) [9] [10] [11] [12], [13] 1 Fig. 1 Flowchart of the proposed system. c 2013 Information

kiyo5_1-masuzawa.indd

DEIM Forum 2009 B4-6, Str

Vol1-CVIM-172 No.7 21/5/ Shan 1) 2 2)3) Yuan 4) Ancuti 5) Agrawal 6) 2.4 Ben-Ezra 7)8) Raskar 9) Image domain Blur image l PSF b / = F(

1,a) 1,b) TUBSTAP TUBSTAP Offering New Benchmark Maps for Turn Based Strategy Game Tomihiro Kimura 1,a) Kokolo Ikeda 1,b) Abstract: Tsume-shogi and Ts

(MIRU2010) Geometric Context Randomized Trees Geometric Context Rand

BOK body of knowledge, BOK BOK BOK 1 CC2001 computing curricula 2001 [1] BOK IT BOK 2008 ITBOK [2] social infomatics SI BOK BOK BOK WikiBOK BO

3807 (3)(2) ,267 1 Fig. 1 Advertisement to the author of a blog. 3 (1) (2) (3) (2) (1) TV 2-0 Adsense (2) Web ) 6) 3

1 UD Fig. 1 Concept of UD tourist information system. 1 ()KDDI UD 7) ) UD c 2010 Information Processing S

[1] SBS [2] SBS Random Forests[3] Random Forests ii

Sobel Canny i

900 GPS GPS DGPS Differential GPS RTK-GPS Real Time Kinematic GPS 2) DGPS RTK-GPS GPS GPS Wi-Fi 3) RFID 4) M-CubITS 5) Wi-Fi PSP PlayStation Portable

独立行政法人情報通信研究機構 Development of the Information Analysis System WISDOM KIDAWARA Yutaka NICT Knowledge Clustered Group researched and developed the infor

3 1 Table 1 1 Feature classification of frames included in a comic magazine Type A Type B Type C Others 81.5% 10.3% 5.0% 3.2% Fig. 1 A co

.,,, [12].,, [13].,,.,, meal[10]., [11], SNS.,., [14].,,.,,.,,,.,,., Cami-log, , [15], A/D (Powerlab ; ), F- (F-150M, ), ( PC ).,, Chart5(ADIns

(a) (b) 1 JavaScript Web Web Web CGI Web Web JavaScript Web mixi facebook SNS Web URL ID Web 1 JavaScript Web 1(a) 1(b) JavaScript & Web Web Web Webji

DEIM Forum 2010 A Web Abstract Classification Method for Revie

情報処理学会研究報告 IPSJ SIG Technical Report Vol.2015-GI-34 No /7/ % Selections of Discarding Mahjong Piece Using Neural Network Matsui

IPSJ SIG Technical Report Vol.2011-MUS-91 No /7/ , 3 1 Design and Implementation on a System for Learning Songs by Presenting Musical St

A Study on Throw Simulation for Baseball Pitching Machine with Rollers and Its Optimization Shinobu SAKAI*5, Yuichiro KITAGAWA, Ryo KANAI and Juhachi

Vol. 42 No MUC-6 6) 90% 2) MUC-6 MET-1 7),8) 7 90% 1 MUC IREX-NE 9) 10),11) 1) MUCMET 12) IREX-NE 13) ARPA 1987 MUC 1992 TREC IREX-N

Honda 3) Fujii 4) 5) Agrawala 6) Osaragi 7) Grabler 8) Web Web c 2010 Information Processing Society of Japan

IPSJ SIG Technical Report Vol.2016-CE-137 No /12/ e β /α α β β / α A judgment method of difficulty of task for a learner using simple

23 Fig. 2: hwmodulev2 3. Reconfigurable HPC 3.1 hw/sw hw/sw hw/sw FPGA PC FPGA PC FPGA HPC FPGA FPGA hw/sw hw/sw hw- Module FPGA hwmodule hw/sw FPGA h

情報処理学会研究報告 IPSJ SIG Technical Report Vol.2013-CVIM-186 No /3/15 EMD 1,a) SIFT. SIFT Bag-of-keypoints. SIFT SIFT.. Earth Mover s Distance

IPSJ SIG Technical Report Vol.2013-CVIM-188 No /9/2 1,a) D. Marr D. Marr 1. (feature-based) (area-based) (Dense Stereo Vision) van der Ma

IPSJ SIG Technical Report Vol.2009-BIO-17 No /5/26 DNA 1 1 DNA DNA DNA DNA Correcting read errors on DNA sequences determined by Pyrosequencing

3_39.dvi

Vol.54 No (Mar. 2013) 1,a) , A Case Study of the Publication of Information on the Japan Earthquake Naoto Matsumoto 1,a

258 5) GPS 1 GPS 6) GPS DP 7) 8) 10) GPS GPS ) GPS Global Positioning System

IPSJ SIG Technical Report Vol.2014-DBS-159 No.6 Vol.2014-IFAT-115 No /8/1 1,a) 1 1 1,, 1. ([1]) ([2], [3]) A B 1 ([4]) 1 Graduate School of Info

21 Key Exchange method for portable terminal with direct input by user

HASC2012corpus HASC Challenge 2010,2011 HASC2011corpus( 116, 4898), HASC2012corpus( 136, 7668) HASC2012corpus HASC2012corpus

(MIRU2008) HOG Histograms of Oriented Gradients (HOG)

IPSJ SIG Technical Report Vol.2010-NL-199 No /11/ treebank ( ) KWIC /MeCab / Morphological and Dependency Structure Annotated Corp

Web ( ) [1] Web Shibboleth SSO Web SSO Web Web Shibboleth SAML IdP(Identity Provider) Web Web (SP:ServiceProvider) ( ) IdP Web Web MRA(Mail Retrieval

it-ken_open.key

The copyright of this material is retained by the Information Processing Society of Japan (IPSJ). The material has been made available on the website

EQUIVALENT TRANSFORMATION TECHNIQUE FOR ISLANDING DETECTION METHODS OF SYNCHRONOUS GENERATOR -REACTIVE POWER PERTURBATION METHODS USING AVR OR SVC- Ju

Transcription:

1,a) 1,b) 1,2,c) 1,d) BtoC EC EC EC EC ImageNet Inception-v3 EC A search system of retrieving images of the similar products without the requirement of the training of feature extractor and the purchase history Yusuke Miyake 1,a) Ryosuke Matsumoto 1,b) Kenji Rikitake 1,2,c) Kentaro Kuribayashi 1,d) Abstract: Electronic Commerce (EC) sites show items of the user s interest following the flow for efficient browsing of the products. For enabling to choose the related products without the activity record of the users, methods using the metadata and the images of the products to suggest a highly accurate choice by deep learning with convolution neural networks have been proposed. Those methods, however, require the large amount of training data and calculation time with a properly structured product metadata, which results in the impediment to the production system deployment. In this report, we propose a search system of retrieving images of the similar products without the requirement of the training of feature extractor and the purchase history, by comparing the similarity with the approximate nearest neighbor search based on the features from the pooling layer before the identification layer, using the Inception-v3 model which claims a good result on ImageNet, an academic benchmark of image classification similar to that of EC site product image characteristics. We implement this system to an EC site and measure the clickthrough rate of the image for each product category to evaluate the effectiveness of directing the user flow by showing the related products by similar images. 1 GMO Pepabo R&D Institute, GMO Pepabo, Inc., Tenjin, Chuo ku, Fukuoka 810-0001 Japan 2 Kenji Rikitake Professional Engineer s Office, Toyonaka City, Osaka 560-0043 Japan a) miyakey@pepabo.com b) matsumotory@pepabo.com c) kenji.rikitake@acm.org d) antipop@pepabo.com c 2017 Information Processing Society of Japan 1

1. BtoC EC [23] EC EC EC Amazon EC [9] EC [2] EC [18] [21] [5], [14] [4] [16] EC [11] [3] EC EC ImageNet Inception-v3 [5] Inception-v3 EC [12] EC [3] EC Inception-v3 EC 2 3 4 5 c 2017 Information Processing Society of Japan 2

2. 2.1 1 [10] EC 2.2 1 EC 2.3 [7] [1] [18] [5], [14] [8], [21] Inception-v3 ILSVRC[14]2014 GoogLeNet[4] ILSVRC2012 1000 Top-5 GoogLeNet 6.67% 3.46% 5.1% *1 [4] [16] [11] [3] EC EC 3 ( 1 ) EC ( 2 ) ( 3 ) *1 http://karpathy.github.io/2014/09/02/what-i-learned-fromcompeting-against-a-convnet-on-imagenet/ c 2017 Information Processing Society of Japan 3

3. 2 2 ( 1 ) EC ( 2 ) 2 EC EC 3.1 2.3 EC ImageNet 2017 4 10 21,841 14,197,122 ILSVRC [14] ImageNet ILSVRC EC ILSVRC Inception-v3 [8], [21] [12] # with tf.gfile.fastgfile(flags.classify_net, rb ) as f: graph_def = tf.graphdef() graph_def.parsefromstring(f.read()) _ = tf.import_graph_def(graph_def, name= ) with tf.session() as sess: 1 Fig. 1 # pool3 = sess.graph.get_tensor_by_name( pool_3:0 ) jpeg_data = tf.placeholder(tf.string) with file_io.fileio(input_csv(), r ) as f: with file_io.fileio(output_csv(), w ) as output: writer = csv.writer(output, lineterminator= \n ) for line in f: product_id, image = image_from_line(line) input_name = DecodeJpeg/contents:0 results = sess.run(pool3, {input_name: image}) features = [ {:.18e}.format(f) for f in results[0][0][0]] writer.writerow([product_id] + features) TensorFlow Python Implementation examples of extracting features using Python and TensorFlow. Inception-v3 Pool3 EC Inception-v3 TensorFlow[19] TensorFlow Python 1 3.2 3.1 EC 3.1 2048 EC Annoy [20] c 2017 Information Processing Society of Japan 4

category_ids.each do c annoy = AnnoyIndex.new(2048) File.foreach("features/features_#{c}.csv") do line Fig. 2 features = line.chomp.split(, ) annoy.add_item(i, features[1..-1].map(&:to_f)) annoy.build(10) annoy.save("ann/category_#{c}.ann") 2 mruby annoy Implementation examples of building database using mruby annoy. Annoy B [17] EC HTTP API API nginx[13] ngx mruby[24] mruby-annoy ngx mruby nginx mruby[22] mruby-annoy Annoy C++ API mruby ngx mruby API mruby-annoy OSS *2 3.1 Annoy mruby-annoy 2 mrubyuserdata ngx mruby mruby init 3 ID ID *2 https://github.com/monochromegane/mruby-annoy userdata = Userdata.new "annoy_data_key" ann_dir = /etc/nginx/conf.d/ann Dir.foreach(ann_dir) do ann next unless ann._with?(.ann ) category_id = ann.split( _ )[1] annoy = AnnoyIndex.new(2048) annoy.load("#{ann_dir}/#{ann}") userdata.s("category_#{category_id}=", annoy) Fig. 3 class NNS def call(env) 3 mruby init Configuration examples of mruby init. params = env[ QUERY_STRING ].split( & ).map { kv kv.split( = ) }.to_h category_id = params[ category_id ].to_i product_id limit = params[ product_id ].to_i = (params[ limit ] 10).to_i userdata = Userdata.new "annoy_data_key" annoy = userdata.s("category_#{category_id}") return not_found unless annoy nns = annoy.get_nns_by_item(product_id, limit) [200, content_type, [nns.to_json]] private def not_found return [404, content_type, def content_type run NNS.new Fig. 4 [{ error => not_found }.to_json]] { Content-Type => application/json;charset=utf-8 } 4 mruby content handler Configuration examples of mruby content handler. % curl http://localhost/nns?category_id=1&product_id=234 [233,234,127,235,223] Fig. 5 5 ngx mruby mruby annoy Execution examples of mruby annoy on ngx mruby. 4 5 c 2017 Information Processing Society of Japan 5

情報処理学会研究報告 4. 実験と考察 提案手法の有効性を検証するため 3 章で構築したデー タベース並びに類似特徴量検索サーバーを EC サイトの プロダクション環境に導入した 導入した EC サイト上で は 商品詳細画面にて関連商品を 12 点表示する欄を設け ており これらの関連商品のクリック数を計測している 本研究では 類似画像検索の有効性を検証するため 今回 提案した手法により選定する商品と EC サイトの既存の 手法で選定する商品間でのクリック率の差を計測した な お EC サイトの既存手法は 独自に運用していた簡易的な ユーザーベースの協調フィルタリングにあたる 具体的に 図 6 対象商品 は 該当商品を評価している他の利用者群を類似ユーザー Fig. 6 Target product. とみなし それぞれの利用者が評価している商品に対して 評価の新しさに重みをつけて一定数取得した後 該当商品 と重複商品を除いたものを関連商品として選定する 実験の手法として まず既存手法により選定した商品を A パターン 提案手法により選定した商品を B パターンと して 両パターンのクリック率を比較検討するため 表示 機会が均等になるよう乱数によりおよそ 50%の確率で表示 のパターンを判定している しかし 3.1 節で構築した時点 以降に登録された商品については 提案手法による選定が 行えないため その場合は既存手法で選定した商品を表示 し 既存手法側のクリック率計測対象とした これは検証 期間中の EC サイトの機会損失を減らす必要から取った措 置である なお 既存手法が協調フィルタリング手法であ る以上 商品に対する情報が蓄積されていない場合は 既 存手法でも選定が行えない場合があるが その場合は 該 当する関連商品なしとして計測の対象から除外している 提案手法の対象となる商品数は約 380 万点であり 検証期 間中の各手法の利用割合は 既存手法 68.57% 提案手法 31.43%であった 図 7 既存手法による選定商品 Fig. 7 Selected products by existing method. また 提案手法のうち 商品特性による有効性の比較を 表 1 クリック率とコンバージョン率の比較 行うため EC サイトに既にあるカテゴリ区分に従い ク Table 1 Comparison of CTR and CVR. リック率を計測して比較 考察を行った 対象のカテゴリ 数は 120 である なお 検証にあたり類似特徴量検索サーバーは これが 参照するデータベースファイルを全てページキャッシュす るのに十分な量のメモリを搭載している必要があった こ れは近似近傍探索の際に木構造を辿る処理がデータベース ファイルの広範囲に渡るランダムアクセスとなり キャッ シュされていない場合に都度ディスクへのアクセスが発生 して 応答時間の遅延が発生し 導入先の EC サイトから の想定要求数を処理できないことが事前の性能評価で確認 されたためである 既存手法 提案手法 クリック率 0.51% 0.76% コンバージョン率 クリック数 0.21% 0.28% 4.1 類似画像検索の評価と考察 対象商品図 6 に対して 既存手法 提案手法によって選 定された関連商品の例をそれぞれ図 7 図 8 に示す 提案 手法において 対象商品の形状や特徴を掴み 類似した商 品を選定できていることがわかる これらの選定した商品 に対するクリック率とコンバージョン率の比較を表 1 に 示す クリック数は閲覧数に対するクリック数から コン バージョン率はクリック数に対する購入回数から求めた 既存手法に比べて提案手法のクリック率が改善しており 提案手法による選定商品が 関連商品としてより興味を引 c 2017 Information Processing Society of Japan 6

2 5 Table 2 Top 5 categories which existing method is effective. / 1.20% 0.40% / 0.76% 0.36% / 0.45% 0.20% / 0.50% 0.28% / 0.56% 0.38% 3 5 Table 3 Top 5 categories which proposed method is effective. 8 Fig. 8 Selected products by proposed method. 4.2 4.1 EC 120 120 28 92 77% 5 2 3 / 0.26% 1.25% 1% / 1.20% 0.40% EC / 0.26% 1.25% / 0.47% 1.46% / 0.53% 1.33% / 0.40% 1.17% / 0.37% 1.07% 5. EC EC EC EC c 2017 Information Processing Society of Japan 7

[15] [1] Andrew McCallum, Information Extraction: Distilling Structured Data from Unstructured Text, Queue - Social Computing Volume 3 Issue 9, pp. 48-57, November 2005 [2] J. Ben Schafer, Joseph Konstan, John Riedl, Recommer systems in e-commerce, Proceedings of the 1st ACM conference on Electronic commerce, pp.158-166, November 03-05, 1999 [3] Bolei Zhou, Agata Lapedriza, Jianxiong Xiao, Antonio Torralba, and Aude Oliva, Learning Deep Features for Scene Recognition using Places Database, In Proc. NIPS, 2014. [4] Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, Andrew Rabinovich, Going Deeper with Convolutions, The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015 [5] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, Zbigniew Wojna, Rethinking the Inception Architecture for Computer Vision, The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2818-2826, 2016 [6] David Goldberg, David Nichols, Brian M. Oki, and Douglas Terry, Using collaborative filtering to weave an information tapestry. Commun. ACM 35, 12, pp. 61-70, December 1992 [7] Dieter Fensel, Ying Ding, Borys Omelayenko, Ellen Schulten, Guy Botquin, Mike Brown, Alan Flett, Product Data Integration in B2B E-Commerce, IEEE Intelligent Systems Volume 16, Issue 4, pp. 54-59, Jul-Aug 2001 [8] K. Fukushima. Neocognitron: a self organizing neural network model for a mechanism of pattern recognition unaffected by shift in position, Biological Cybernetics, 36(4):93202, 1980 [9] Greg Linden, Brent Smith, Jeremy York, Amazon.com Recommations: Item-to-Item Collaborative Filtering, IEEE Internet Computing Volume 7 Issue 1, pp. 76-80, January 2003 [10] Hyung Jun Ahn, A new similarity measure for collaborative filtering to alleviate the new user coldstarting problem, Information Sciences 178, pp. 37-51, 2008 [11] Jeff Donahue, Yangqing Jia, Oriol Vinyals, Judy Hoffman, Ning Zhang, Eric Tzeng, Trevor Darrell, DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition. In Proc. ICML, 2014 [12] Matthew D. Zeiler, Rob Fergus, Visualizing and Understanding Convolutional Networks, In Proc, Computer VisionECCV, pp. 818-833, 2014 [13] Nginx, Nginx, http://nginx.org/ja/. [14] Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. 2015. ImageNet Large Scale Visual Recognition Challenge. Int. J. Comput. Vision 115, 3, pp. 211-252, December 2015 [15] Paolo Ciaccia, Marco Patella, and Pavel Zezula, M-tree: An Efficient Access Method for Similarity Search in Metric Spaces, In Proceedings of the 23rd International Conference on Very Large Data Bases (VLDB 97), pp. 426-435, 1997 [16] Pulkit Agrawal, Ross Girshick, Jitra Malik, Analyzing the Performance of Multilayer Neural Networks for Object Recognition, In Proc. ECCV, 2014 [17] Sanjoy Dasgupta, Kaushik Sinha, Randomized partition trees for exact nearest neighbor search, JMLR: Workshop and Conference Proceedings vol 30, 121, 2013 [18] Shinzato Keiji, Satoshi Sekine, Unsupervised Extraction of Attributes and Their Values from Product Description., IJCNLP, 2013 [19] Martn Abadi, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Ian Goodfellow, Andrew Harp, Geoffrey Irving, Michael Isard, Yangqing Jia, Rafal Jozefowicz, Lukasz Kaiser, Manjunath Kudlur, Josh Levenberg, Dan Mane, Rajat Monga, Sherry Moore, Derek Murray, Chris Olah, Mike Schuster, Jonathon Shlens, Benoit Steiner, Ilya Sutskever, Kunal Talwar, Paul Tucker, Vincent Vanhoucke, Vijay Vasudevan, Fernanda Viegas, Oriol Vinyals, Pete Warden, Martin Wattenberg, Martin Wicke, Yuan Yu, Xiaoqiang Zheng, TensorFlow: Large- Scale Machine Learning on Heterogeneous Distributed Systems, 2015. Software available from tensorflow.org [20] Wen Li, Ying Zhang, Yifang Sun, Wei Wang, Wenjie Zhang, Xuemin Lin, Approximate Nearest Neighbor Search on High Dimensional Data Experiments, Analyses, and Improvement (v1.0), In Proc. 2016 [21] Y. Lecun, L. Bottou, Y. Bengio, P. Haffner, Gradientbased learning applied to document recognition, Proc. IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998 [22] NPO Ruby, http://forum.mruby. org/. [23], 27, 2016 [24],, mod mruby Web,, Vol.55, No.11, pp.2451-2460, 2014 11 [25],, CVIM, 2009-CVIM-169(13), 1-12, 2009-11-19 c 2017 Information Processing Society of Japan 8