4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information).,,,,.,

(JBE Vol. 20, No. 6, November 2015) (Special Paper) 20 6, 2015 11 (JBE Vol. 20, No. 6, November 2015) http://dx.doi.org/10.5909/jbe.2015.20.6.848 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) a), a), a), a), a) Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information Jongbin Park a), Han-Duck Lee a), Kyung-Won Kim a), Jong-Jin Jung a), and Tae-Beom Lim a). IT.,..... Abstract In this paper, we introduce a new web-based PR video making service system. Many video editing tools have required tough editing skill or scenario planning stage for a just simple PR video making. Some users may prefer a simple and fast way than sophisticated and complex functionality. To solve this problem, it is important to provide easy user interface and intelligent classification and recommendation scheme. Therefore, we propose a new template classification and recommendation scheme using a topic modeling method. The proposed scheme has the big advantage of being able to handle the unstructured meta data as well as structured one. Keyword : Video, Classification, Recommendation, Topic, Modeling, Scene, Template

4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information).,,,,., [1].,,,,.. TV.,, TV.,.,. [2-4]. a) (Korea Electronics Technology Institute) Corresponding Author : (Jongbin Park) E-mail: jpark@keti.re.kr Tel: +82-2-6388-6699 ORCID: orcid.org/0000-0002-6123-5300 (IITP) SW. [ (B0101-15- 0559), SW ] 2015. Manuscript received September 15, 2015; Revised October 29, 2015; Accepted October 29, 2015..,..,,. [5],[6] (Needs). [6]. IT..,,..,..

(JBE Vol. 20, No. 6, November 2015) (Topic Modeling). II, III. IV, V.. 1. 1.,,. 1. [6] Fig. 1. PR video making service system [6], IT UI.... 2. MLT.,, 2 3..,. API. API. [5],.. [6] MLT(Media Lovin' Toolkit) [7]. MLT

4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) [7]. API,,,,. MLT LGPL. [7]. MLT.. 1. 2. ( Gathering of resources ),,. MLT, CMS(Content Management Server)., (1) (Scene template),.. (2). (Text String). (3).,,,. (4) (5).., MLT XML. (4).. 2. Fig. 2. Flowchart of PR video making using strategic meta information

(JBE Vol. 20, No. 6, November 2015), (1) 2 Strategic meta information,. 1 (2). ( ), ( ), ( ), ( :, ), CMS(Content Manage- ment Server). (Source). 1. DB XML, JSON [8],[9].. 1 1. Table 1. Examples of strategic meta information ( ) ( ) ( ) : 800, : 1,000, 50 50 { : 50, :, :, : } 1 { :,, : },, { :, : } (UI). { :, : } 30%, 12%,...,, { :, :,, : } Table 1. Examples of strategic meta information Source (Example) Internal processing (Example) Non-structured format (Example) Structured format Marketing Strategy Promotional Strategy User Information User Intention Feedback Information Current monthly sales: 800KRW, Target monthly sales: 1,000KRW, Credit card usage information Primary processing of marketing strategy and trend information Analysis of type of business or location information Estimation of user intention using user interface Analysis of feedback information from streaming, download, comment, etc please, make a ad video for digital signage intended for fifties women please, make a ad video in order to emphasize the calm and warm atmosphere cafe, Seongnam-si, Gyeonggi-do please, make a ad video. Americano is a kind of loss leader. I want to get a profit from the fruit juices. humor contents 30%, fitness contents 12%,..., through the web blog, playing mostly on smart-phone { age : fifties, gender of potential customers : female, medium of advertisement : outside digital signage, type of PR : video } { atmosphere : calm, warm, type of PR : video } { type of business : cafe, location : Seongnam-si, Gyeonggi-do } { strategic item 1 : Americano coffee, strategic item 2 : fruit juice } { access device : smart-phone, type of thema : humor, fitness, traffic source : web blog. }

4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) (6). 2. (Topic Modeling) 2 Recommendation and categorization of scene templates using topic modeling.,.,,. (tag) (Vector Space), TF-IDF(Term Frequency-Inverse Document Frequency) [10]. TF-IDF, (Search Keyword). (word or term) [11]. (Topic modeling).. (1) (document) (topic, theme), (2) (annotation), (3) (organization), (summarization), (searching), (prediction) [11]. (Clustering) (Machine Learning) (Unsupervised Learning). LSA(Latent Sematic Analysis/Indexing) [12] LDA(Latent Dirichlet Allocation) [11]. LSA (PCA; Principle Component Analysis). (7) SVD(Singular Value Decomposition). (7), (word or term),. SVD,, (Data Compression).,. (Eigenvalue) (Diagonal matrix) (Strength) [12]. LDA (topic), (document), (word)...,,, [11]. LDA (parameter).. bag-of-words. EM(Expectation Maximization). LDA

(JBE Vol. 20, No. 6, November 2015) (Dirichlet distribution), [11]. 3. LSA. TF-IDF, LDA. (2) ( ). 2. 2. Table 2. Update Dictionary and Topic Modeling ( 1) ( 2) ( 3) ( 4) ( 5) ( 6) (6) (document).. (Dictionary).,. LSA (update).. 1 4. (Basis vector) (projection). Table 2. Update Dictionary and Topic Modeling (Step 1) Make a combined strategic meta string as (6) (Step 2) Dictionary update after word extraction (Step 3) Save the meta document after conversion of meta information (Step 4) Update topic information using LSA with topics (Step 5) If a new template is added, repeat the steps from 1 to 4 (Step 6) Calculate a similarity [11],[12]. 3., (similarity) (8) (Cosine similarity)., 1, 0, 1. cos 3. Table 3. Scene Template Recommendation using Strategic Meta Information ( 1) ( 2) ( 3) ( 4) ( 5),, (6). (Query statement). 1 2.. 2. (8). (Sorting).

박종빈 외 4인 : 홍보동영상 제작 서비스를 위한 전략메타정보 기반 장면템플릿 분류 및 추천 (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) Ⅳ. 개발결과 Table 3. Scene Template Recommendation using Strategic Meta Information (Step 1) (Step 2) (Step 3) (Step 4) (Step 5) Make a query statement using strategic meta information Extract a word list from query statement using dictionary Projection into k-dimensional topic space of the query statement Calculate similarity between query statement and K-dimensional topic space as (8) Sort and print scene-templates as the similarity order. 초기화면 (a) (a) Introduction Page 홍보의도 선택 (d) (d) User Intention Choice Page 가이드 사용자 환경의 예 (g) 2 (g) Example of Guide User Interface 2 1. 홍보동영상 제작 서비스 시스템 개발 내용 제안하는 홍보동영상 제작 서비스 시스템은 웹상에서 서 베 연동을 통한 동적 서비스 제공을 위해 PHP언어를 사용해서 구현했다. 하드웨어 시스템은 사용자 정보를 관리하는 프론트페이지서버(Front Page Server), 동영상 렌더링을 담당하는 렌더링서버(Rend비스를 제공한다. 데이터 이스 콘텐츠 관리 추천 템플릿 (e) (e) Recommended Templates 진행 페이지 및 문자발송 설정 (h) (h) Progress and Short Message Service Setting Fig. 3. PR Video Making Service System and User Interface [13] ering Server), 만들어진 영상을 저장하고 스트리밍하는 (b) (b) Contents Management 그림 3. 제작된 홍보동영상 제작 서비스 시스템 및 사용자 환경 855 홍보물 관리 (c) (c) PR Video Management 가이드 사용자 환경의 예 (f) 1 (f) Example of Guide User Interface 1 영상제작 완료 (i) (i) Completion of PR Video Making

(JBE Vol. 20, No. 6, November 2015) CMS (Contents Management Server). 3. (a) (b). (c). (d), (e)., (Guide User Interface).. (f), (g). CPU 3GHz (Single thread) 3 ~20. (Short Message Service) (h).. 4, UserAgent, IP. 4. Fig. 4. User Interface for System Administrator

4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) 2. 3(e). [14] 40. 4. 7~10. {,,,,,, } 7. { } 5. 4 0~4 { }, 5~9 { }. 4. Table 4. Examples of meta information for each scene templates [0],,,,,,,, (travel, myself, illustration, photo, travel photo, goods, rising-sun, mountain, sea) [1],,,,,,,,,,, (travel, diary, memory, photo, video, cute, epilogue, event, propose, photo, video, restaurant) [2],,,,,,,, (travel, outdoor, mountain, extreme, sports, sea, accommodations, condominium, pension) [3],,,,,,,, (travel, usa, europe, southeast-asia, schedule, weather, sea, mountain, location) [4],,,,,,,, (travel, relax, rest, mild, weather, love, happy, feeling, comfortable) [5], 1 2,,,,, (accommodation, two-days-and-one-night, individual, image, package, fresh, travel) [6],,,,,, (accommodation, travel-place, cafe, real-estate, luxury, signage, postcard) [7],,,,, (accommodation, premium, portfolio, real-estate, loan, bank) [8],,,,,, (accommodation, hotel, home-stay, price, low-price, comfortable, traffic) [9],,,,,, (accommodation, travel, warm, travel-place, low-price, home-stay, air-bnb) [10],,,,,,, (restaurant, snack, taste, health, traffic, low-price, school, MSG) [11],,,,,,,, (restaurant, japanese-food, taste, fresh, material, tuna, odoro, sushi, udon) [12],,,,,,,,,, (restaurant, korean-food, traditional, traditional-korean-food, taste, clean, dongchimi, siregi, material, MSG) [13],,,,,,,,, (restaurant, korean-food, menu, bibimbap, dongchimi, rice, clean, high-class, taste) [14],,,,,,,,,, (restaurant, jjyukkumi, squid, small-octopus, taste, fresh, material, spicy, sour, sweet, MSG) [15],,,,,, (romantic, lyrical, cafe, flower-shop, book-store, PR-video, video) [16],,,,,,, (cafe, weather, calm, warm, flavor, simple, daily, memory) [17],,,,,, (cafe, lively, happy, live, image, lyrical, template) [18],,,,, (cafe, mild, music, background, interior, menu) [19],,,,, (cafe, coffee, dessert, interior, goods, menu) [20],,,,,,,, (design, composition, simple, spatial, message, implication, ending, opening, deep-blue) [21],,,,, (design, simple, PR-movie, spatial, deep-blue, ending) [22],,,,,,,,,, (design, gold, pop-up, easy, simple, core, topic, attractiveness, implication, ending, opening) [23],,,,,,,,,,,,,,, (design, wooden, eyes, wood, goods, festival, restaurant, service, dynamic, core, topic, attractiveness, implication, ending, opening, moderation) [24],,,,,,,,,,, (design, white-design, eyes, white, beauty, hospital, restaurant, service, attractiveness, implication, ending, opening) [25] SF,,,,,,, (SF, game, PR, dreamlike, dynamic, future, world, adventure) [26] SF,,,,,, (SF, dreamlike, game, PR, image, photo, making) [27],,, RPG,, (game, trailer, large-scale, RPG, sequence, character) [28],,,,,,,,, (game, mobile, image, logo, play, image, casual, casuual, thema, version) [29],,,,, (casual, casuual, video, android, landscape, thema) [30],,,, (pure, mild, cosmetics, goods, ad) [31],,, (vintage, old-film, film, feeling) [32],,,,,, (light-stream, flare, moving, spatial, fancy, electron, fast) [33],,,, (relax, text, sensitivity, sound, variety) [34],,,,,, (happy, daily, sensitive, font, calm, noise, sensitivity) [35],,,,,,,,, (goods, PR, slide-show, slide, cafe, restaurant, travel, real-estate, ) [36],, (newness, fresh, smoothie) [37],,,,,,,, (humor, smile, happy, satisfaction, curious, ha-ha-ha, ho-ho-ho, smiling) [38],,,,,,,,, (love, happy, satisfaction, with, everybody, smile, relax, vanilla, sky, nature) [39],,,,,,,,,, (wait, memory, her, luv, tired, loneliness, heart, pain, alone, thought)

(JBE Vol. 20, No. 6, November 2015) { } 36. 5 4 LSA,. gensim [15]. 5 4 10 1, 12, 14, 11, 13, 8,.... (8).. 5 TF-IDF LSA 40., 5 TF-IDF LSA. TF-IDF LSA 0 1 5. LSA Table 5. Example of topic modeling using LSA * : * Example of strategic meta information { :,, :,, : } { Type of business : snack-bar, restaurant, pr strategy : near school, easy accessibility, low-price, pr intension : delicious food without MSG } * : ( 4 10 ) * Extracted word list : (The same as 10 th meta information of dictionary) {,,,,,,, } {restaurant, snack-bar, taste, health, traffic, low-price, school, MSG} * 10 (K=10) * Example of topic modeling results with K=10 [0] -0.393*"" + -0.393*" " + -0.351*" " + -0.351*"" + -0.282*"" + -0.204*" " + -0.199*" " + -0.199*"" + -0.199*"" + -0.191*" [1] 0.401*" " + 0.337*" " + 0.335*" " + 0.216*" " + 0.210*"" + 0.194*" " + 0.173*" " + 0.153*" " + 0.151*" " + 0.151*" [2] -0.396*" " + 0.302*" " + -0.286*" " + 0.278*" " + -0.210*"" + 0.200*" " + 0.191*"" + -0.164*" " + -0.164*" " + -0.156*" [3] -0.501*" " + -0.372*" " + -0.222*" " + -0.222*"" + -0.222*"" + 0.181*" " + -0.170*"SF" + -0.170*"" + -0.169*" " + -0.169*" [4] -0.379*"" + 0.373*" " + -0.368*" " + 0.257*" " + 0.186*"" + -0.177*" " + -0.177*" " + -0.177*" " + -0.177*"" + -0.177*" [5] -0.414*" " + -0.324*" " + -0.262*" " + -0.262*" " + 0.171*" " + -0.150*" " + -0.150*" " + -0.150*"" + -0.150*" " + -0.150*" [6] 0.485*" " + 0.199*" " + 0.196*" " + -0.191*" " + -0.191*" " + 0.172*" " + -0.167*" " + -0.157*" " + 0.150*"" + 0.144*" [7] -0.239*" " + 0.216*" " + -0.194*" " + 0.175*" " + 0.175*" " + 0.154*" " + 0.146*" " + -0.142*" " + -0.141*"" + -0.141*" [8] 0.406*" " + -0.208*" " + -0.198*" " + -0.178*" " + -0.178*" " + 0.174*"" + 0.158*"" + -0.151*" " + 0.149*"" + -0.138*" [9] -0.217*" " + -0.212*" " + -0.212*"" + -0.212*" " + -0.212*" " + -0.212*" " + -0.212*" " + -0.212*" " + -0.212*"" + -0.212*" [0] -0.393*"design" + -0.393*"ending" + -0.351*"implication" + -0.351*"opening" + -0.282*"attention" + -0.204*"simple" + -0.199*"eyes" + -0.199*"restaurant" + -0.199*"service" + -0.191*"topic [1] 0.401*"restaurant" + 0.337*"taste" + 0.335*"travel" + 0.216*"material" + 0.210*"MSG" + 0.194*"photo" + 0.173*"cafe" + 0.153*"accommodation" + 0.151*"korean-food" + 0.151*"neat [2] -0.396*"travel" + 0.302*"taste" + -0.286*"photo" + 0.278*"restaurant" + -0.210*"video" + 0.200*"material" + 0.191*"MSG" + -0.164*"mountain" + -0.164*"sea" + -0.156*"accommodation [3] -0.501*"image" + -0.372*"game" + -0.222*"thema" + -0.222*"casual" + -0.222*"casuual" + 0.181*"travel" + -0.170*"SF" + -0.170*"dreamlike" + -0.169*"version" + -0.169*"play [4] -0.379*"video" + 0.373*"cafe" + -0.368*"photo" + 0.257*"accommodation" + 0.186*"real-estate" + -0.177*"propose" + -0.177*"diary" + -0.177*"epilogue" + -0.177*"event" + -0.177*"cute [5] -0.414*"happy" + -0.324*"love" + -0.262*"smile" + -0.262*"satisfaction" + 0.171*"accommodation" + -0.150*"everybody" + -0.150*"relaxed" + -0.150*"vanilla" + -0.150*"nature" + -0.150*"sky [6] 0.485*"cafe" + 0.199*"menu" + 0.196*"memory" + -0.191*"sea" + -0.191*"mountain" + 0.172*"interior" + -0.167*"travel" + -0.157*"accommodation" + 0.150*"video" + 0.144*"simple [7] -0.239*"simple" + 0.216*"PR" + -0.194*"memory" + 0.175*"smile" + 0.175*"satisfaction" + 0.154*"goods" + 0.146*"photo" + -0.142*"weather" + -0.141*"loneliness" + -0.141*"luv [8] 0.406*"accommodation" + -0.208*"weather" + -0.198*"simple" + -0.178*"mountain" + -0.178*"sea" + 0.174*"travel-place" + 0.158*"real-estate" + -0.151*"PR" + 0.149*"video" + -0.138*"game [9] -0.217*"PR" + -0.212*"luv" + -0.212*"waiting" + -0.212*"mind" + -0.212*"thought" + -0.212*"suffering" + -0.212*"alone" + -0.212*"Mr" + -0.212*"loneliness" + -0.212*"tired *, 40 ( ) * Recommendation results in 40 templates (Sorted with high similarity order) [(10, 0.99999994), (12, 0.96954387), (14, 0.95164019), (11, 0.9509294), (13, 0.9145661), (8, 0.35120615), (35, 0.19930224), (7, 0.17918736), (9, 0.16030958), (6, 0.15088126), (5, 0.14760816), (2, 0.086455025), (29, 0.074984722), (39, 0.049634695), (28, 0.047898445), (23, 0.02247897), (24, 0.019126773), (38, 0.0090787606), (31, 0.0), (32, 0.0), (36, 0.0), (37, -0.00530951), (1, -0.0074476823), (18, -0.012554379), (22, -0.01362268), (4, -0.014859326), (17, -0.016441755), (19, -0.0207964), (3, -0.021926962), (30, -0.024547208), (20, -0.030332856), (0, -0.037311688), (21, -0.048699819), (33, -0.060024895), (34, -0.074627325), (26, -0.079447947), (27, -0.080461517), (16, -0.08578527), (15, -0.086388439), (25, -0.11432723)]

4 : (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) 5(a) 100% 0. LSA 5. 1 0. LSA TF-IDF. 5 LDA. 3. 3(d) (6) (Query statement). 3(e). 3(e) ( ), (a) TF-IDF (b) LSA, =1 (c) LSA, =5 (d) LSA, =10 (e) LSA, =15 (f) LSA, =20 (g) LSA, =25 (h) LSA, =30 (i) LSA, =35 5. Fig. 5. Results of scene templates

(JBE Vol. 20, No. 6, November 2015) 30,... LSA 5.......,. (References) [ 1] https://en.wikipedia.org/wiki/public_relations [ 2] E. Jung, J. Choi, Y. Shon, "Influence of advertising type and gender on advertising effectiveness: The efficacy of storytelling advertising", Korean Journal of the science of Emotion & sensibility, pp.489-502, 2012 [ 3] Michael Miller, YouTube for business: Online video marketing for any business, 2nd Edition, Pearson Education, 2011 [ 4] David Meerman Scott, The new rules of marketing and PR: how to use social media, blogs, news releases, online video, and viral marketing to reach buyers directly, John Wiley & Sons, 2009 [ 5] J. Park, M. Park, K. Kim, J. Jeong, J. Lee, and T. Lim, Development of Cloud-based Service Framework for Promotional Video Synthesis, KICS, 2015 Winter Conference, pp. 795-796, 2015 [ 6] J. Park, K. Kim, J. Jeong, and T. Lim, "Development of Processing Interface for Promotional Video Making System based on MLT Multimedia Framework", 2015 Korean Society of Broadcast Engineers Summer Conference, w2.4-4, pp. 1-2, 2015 [ 7] MLT Multimedia Framework, http://www.mltframework.org/ [ 8] Segaran, Toby. Programming collective intelligence: building smart web 2.0 applications. " O'Reilly Media, Inc.", 2007 [ 9] McKinney, Wes. Python for data analysis: Data wrangling with Pandas, NumPy, and IPython. " O'Reilly Media, Inc.", 2012 [10] G. Salton and M. McGill, editors. Introduction to Modern Information Retrieval. McGraw-Hill, 1983 [11] Blei, David M., Andrew Y. Ng, and Michael I. Jordan. "Latent dirichlet allocation." the Journal of machine Learning research 3 (2003): 993-1022 [12] Deerwester, Scott C., et al. "Indexing by latent semantic analysis." JAsIs 41.6 pp.391-407, 1990 [13] PHP, http://www.w3schools.com/php/ [14] Shakr, https://www.shakr.com/ [15] gensim, https://radimrehurek.com/gensim/

박종빈 외 4인 : 홍보동영상 제작 서비스를 위한 전략메타정보 기반 장면템플릿 분류 및 추천 (Jongbin Park et al.: Classification and Recommendation of Scene Templates for PR Video Making Service based on Strategic Meta Information) 저자소개 박종빈 - 년 : 성균관대학교 정보통신공학부 학사 년 : 성균관대학교 전자전기컴퓨터공학부 석사 년 : 캐나다통신연구소(CRC) 방문연구원 년 : 성균관대학교 전자전기컴퓨터공학부 박사 년 : 성균관대학교 정보통신공학부 박사후 연구원 년 ~ 현재 : 전자부품연구원 스마트미디어연구센터 : orcid.org/0000-0002-6123-5300 주관심분야 : 멀티미디어신호처리, 데이터 압축, 정보추론, IoT 및 임베디드 시스템 기술 2004 2006 2008 2011 2012 2012 ORCID 이한덕 - 년 : 건국대학교 학사 년 : 건국대학교 석사 년 ~ 현재 : 전자부품연구원 스마트미디어연구센터 : orcid.org/0000-0001-9474-4171 주관심분야 : 비디오 인코딩, 인공지능, 병렬처리, 분산처리 2008 2011 2011 ORCID 김경원 - 년 : 한국외국어대학교 컴퓨터공학과 학사 년 : 한국외국어대학교 컴퓨터공학과 석사 년 : 건국대학교 컴퓨터 정보통신공학과 박사수료 년 ~ 현재 : 전자부품연구원 스마트미디어연구센터 : orcid.org/0000-0001-6530-8426 주관심분야 : 메타데이터, 스마트 TV, N-스크린 서비스, 맞춤형방송, 멀티미디어 검색 2001 2003 2013 2004 ORCID 정종진 - 년 : 성균관대학교 정보통신공학부 학사 년 : 성균관대학교 전자전기컴퓨터공학부 석사 년 ~ 현재 : 전자부품연구원 스마트미디어연구센터 : orcid.org/0000-0003-3924-8938 주관심분야 : 멀티미디어신호처리, 빅데이터 기반 스마트 서비스, 클라우드 기반 스마트홈 플랫폼 1997 2002 2002 ORCID 임태범 - 년 : 서강대학교 물리학과 학사 년 : 서강대학교 전자계산학과 석사 년 : 대우전자 영상연구소 전임연구원 년 : 건국대학교 컴퓨터공학과 박사 년 ~ 현재 : 전자부품연구원 스마트미디어연구센터장 : orcid.org/0000-0003-1173-6606 주관심분야 : 홈네트워크 솔루션, IoT, 클라우드 방송, DTV방송, 맞춤형방송, 멀티미디어 검색, IPTV 1995 1997 2002 2012 2002 ORCID 861