SNS. Selection of news story series based on the responses by SNS users for the summarization of multiple news videos

Transcription

1 THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS TECHNICAL REPORT OF IEICE. SNS, SNS SNS SNS SNS Selection of news story series based on the responses by SNS users for the summarization of multiple news videos Kosuke KATO, Ichiro IDE, Daisuke DEGUCHI,, and Hiroshi MURASE Nagoya University, Graduate School of Information Science Furo-cho, Chikusa-ku, Nagoya-shi, Aichi, Japan Nagoya University, Information and Communications Headquarters Furo-cho, Chikusa-ku, Nagoya-shi, Aichi, Japan Abstract Archiving news videos is regarded as important since they are valuable as sources of important social information. When exploiting archived news videos as information sources, it is important to track the flow of topics to understand their contents comprehensively. Concerning it, a method that structures the chronological semantic relations between news stories, namely the topic thread structure has been proposed. It allows the understanding of the circumstance of the topic by tracking related stories one after another from the initial story. However, it imposes a user to watch many stories along the topic thread structure when it contains various topics. On the other hand, social networking services (SNS) are explosively prevalent. SNS users aim to send and receive information in which they are interested. Thus, we considered editing news videos from the viewpoint of the viewers, and we have decided to summarize multiple news videos based on the responses by SNS users. In this report, we introduce a method that selects news story series which will be used for the news video summarization. Key words News video, news video archive, SNS 1

2 1 : mediawalker 1. [1] Duygulu [2] [3] Wu [4] 2 [5] mediawalker [6] mediawalker 1 SNS SNS [7] SNS SNS SNS Kobayashi [8] Doman [9] Kobayashi Twitter Doman Twitter SNS 2. SNS SNS [5] CC w 2

3 2 w 1 10 w T r T S (i) T S (j) : T S (j) T S (i) T S (i) : T S (j) T S (i) T S (i) T S (j) 2. 2 [5] (1) C 0 = S O S = S O (2) S S c (j)(j = 1,... J) C 0 S c (j) S c (j) C 0 = S c (j) (3) S 2 (4) 2. 3 Web [10] SNS Web Web SNS Web Web CC Web (1) (2) (3) (4) 2. 4 CC SNS (1) CC (2) tf-idf (3) CC W CC Web W article Web SNS N i j W story (i, j) { W story (i, j) = W CC (i, j) N (if W article (i, j) = 0) (1) W story (i, j) = W CC (i, j) (otherwise) W CC (i, j) i j tf-idf W article (i, j) i Web j k j W cluster (k, j) W cluster (k, j) = W story (i, j) (2) i c k c k k

4 2013/12/05#6 2013/12/06#2 2013/12/09#8 2013/12/10#6 2013/12/18#8 2013/12/31#3 2014/01/09#8 2014/01/06#6 yyyy/mm/dd#n 2014/01/07# /01/08# /01/09# ,080 3 CC SNS SNS NHK 7 Web NHK NEWSWEB SNS Facebook SNS (1) N = SNS a SNS b SNS , , , , , , , , , , a SNS b SNS , , ,

5 4 3 a SNS b SNS , , , , , , , , , , ,296 36,960 34,750 31,448 23,843 22,547 19,238 16,342 15,922 15,393 13,029 11,700 11,537 11,072 9,484 9,072 8,935 8,255 8,136 7, a 1 b 2 c 3 46,296 2,472 7,550 36,960 1,552 5,917 34,750 1,342 5,234 31, ,775 23, ,328 20, ,078 18, ,352 16, ,343 15, ,309 15, ,272 12, ,222 11, ,140 11, ,130 11, ,098 9, ,095 9, ,041 8, , , , t t =

6 7 2013/12/05#6 16, /12/05#6 24, /12/05#6 13, /12/06#2 12, /12/06#2 14, /12/06#2 18, /12/06#2 18, /12/06#2 11, /12/06#2 12, /12/18#8 12, /12/18#8 13, SNS Web SNS tf-idf SNS [6] Ichiro Ide, Tomoyoshi Kinoshita, Tomokazu Takahashi, Shin ichi Satoh, and Hiroshi Murase, mediawalker: A video archive explorer based on time-series semantic structure, Proc. 15th ACM Int. Conf. on Multimedia, pp , Sept [7] 23 h23/html/nc html [2014/8/3 ] [8] Takashi Kobayashi, Tomokazu Takahashi, Daisuke Deguchi, Ichiro Ide, and Hiroshi Murase, Detection of biased broadcast sports video highlights by attribute-based tweets analysis, Proc. 19th Int. Conf. on Multimedia Modeling, Lecture Note in Computer Science, Vol.7733, pp , Jan [9] Keisuke Doman, Taishi Tomita, Ichiro Ide, Daisuke Deguchi, and Hiroshi Murase, Event detection based on Twitter Enthusiasm Degree for generating a sports highlight video, 22nd ACM Int. Multimedia Conf. (ACM MM2014), Nov (To appear). [10],,,, SNS Web MVE , Mar [1] Vol.64 No.3 pp Mar 2010 [2] Pinar Duygulu, Jia-Yu Pan, and David A. Forsyth, Towards auto-documentary: Tracking the evolution of news stories, Proc. 12th ACM Int. Conf. on Multimedia, pp , Oct [3] Jonathan G. Fiscus and George R. Doddington, Topic detection and tracking evaluation overview, In James Allan, ed Topic detection and tracking: Event-based information organization, Kluwer Academic Publishers, pp.17 32, [4] Xiao Wu, Chong-Wah Ngo, and Qing Li, Threading and auto documenting news videos, IEEE Signal Processing Mag., Vol.23, No.2, pp.59 68, Mar [5] Vol.23 No.5 pp Sept