1,a) 1 1 1 2014 9 20, 2015 1 5 TPO Extracting Purpose-for-Action to Enhance Local Information Service Noriko Yokoyama 1,a) Kaname Funakoshi 1 Hiroyuki Toda 1 Yoshimasa Koike 1 Received: September 20, 2014, Accepted: January 5, 2015 Abstract: To suggest a target action and encourage the user to perform it, motivation and ability are necessary when a trigger is offered. However, traditional local information services focus on giving information related to the ability to perform the behavior (cost, place, etc.) and motivation is not considered; without effective motivation the target action is unlikely to be performed. Our goal is to identify an effective motivation approach. To achieve this goal, we collect purpose-for-action. In this paper, we propose a method to extract purpose-for-action from social media texts by using clue expressions and modification structure. Moreover, we conduct a user experiment whose results confirm that showing purpose-for-action yields effective motivation. Keywords: text mining, social media, purpose-for-action extraction, action support 1. *1 *2 *3 Foursquare *4 1 1 NTT NTT Service Evolution Laboratories, Yokosuka, Kanagawa 230 0847, Japan a) yokoyama.noriko@lab.ntt.co.jp *1 http://tabelog.com/ *2 http://www.jalan.net/kankou/ *3 http://local.smt.docomo.ne.jp/g.php *4 Foursquare https://ja.foursquare.com/ c 2015 Information Processing Society of Japan 17
1 Fogg Fogg Behavior Model [1] 3 1 2 1 1 2 2 2 1 1 information cascade [2] TPO [3] 2 3 2 1 2 2 1 2 2 Fig. 2 Target area of this paper. 1 Fig. 1 Fogg behavior model Fogg behavior model. 3 Fig. 3 Example of application image. c 2015 Information Processing Society of Japan 18
4 Fig. 4 Extracting purpose-for-action from experience information. 1 4 2 3 4 5 6 7 8 2. 2.1 2 (1) (2) (1) (2) (1) [4] [5] twitter LDA (2) [6], [7], [8], [9] [9] (1) (2) 2 (2) (2) (1) (2) 2.2 Foursquare information cascade Google *5 Amazon *6 [10] *5 https://www.google.co.jp/ *6 http://www.amazon.com/ c 2015 Information Processing Society of Japan 19
3. 5 1 1 2 5 Fig. 5 Definition of purpose-for-action. 4. 4.1 Q&A 1 4.2 [9] 10 100 10 [11] 10 [12] 12 12 430 1 2 (1) (2) 1 (1) (2) (3) c 2015 Information Processing Society of Japan 20
1 (1) (2) (1) (1) 1 15 2013 2 1 1 100 10 5 3 2 0 1 Table 1 Clue expressions extracted from blogs. 2 Table 2 Frequency of clue expressions. 5. 2 5.1 [6] [6] 6 4 1. 2. Jdep [13] 3. 1 1 2 1 2 2 4. 3 1 1 2 1 2 2 3 4 1 1 6 Fig. 6 Purpose-for-action extraction using clue expression and modification structure. c 2015 Information Processing Society of Japan 21
1 2 2 6 2.1 6.1 5.2 5.1 1. 2. 3. 1 4. 2 6. F 2013 2 1 5 [13] 5.1 4 2 F 5 F = = 2 F = + BL2 BL1 2 BL2 3 4 5 5 2 Table 3 3 The number of correct examples. 6.1 5 5.1 2 1 3 2 BL1 [13] 5.1 4 BL2 Table 4 4 The number of extracted examples. c 2015 Information Processing Society of Japan 22
5 Table 5 Extraction accuracy. Table 6 6 Initial improvement policies. BL1 1 F 2 2 FP: False Positive2 FP (1) (2) (3) (4) / (5) 5.1 FP 6 1. 2. 2 7 Table 7 The change in the number of errors (close test). 8 Table 8 Removal of errors (open test). 7 62% 78% F 67% 70% 2014 2 1 2 BL2 BL2 BL2 F 8 F BL2 2 F 8 c 2015 Information Processing Society of Japan 23
9 Table 9 Clustering. 10 2 Table 10 Example of question 2. 6.2 6.1 FP 6.1 2 7 6.1 P@Nc 2 2 9 F 7. 6 1 7.1 1 2 ex. 5 10 (1) (2) (1) 6 5 11 [11] 5 / // / 2013 1 2013 12 1 4 6.1 6.2 1 [14] 50 18 50 36 1 400 7.2 or x 6 11 1 x > 0 342 85% c 2015 Information Processing Society of Japan 24
11 Table 11 Evaluation of the effect of system. 5W1H 2 12 Table 12 Examples of purpose-for-action scores (Running). + 7% 75% 12 13 36% [1] Fogg, B.J.: A Behavior Model for Persuasive Design, Proc. 4th International Conference on Persuasive Technology, Article No.40 (2009). [2] Bikhchandani, S., Hirshleifer, D. and Welch, I.: Learning from the Behavior of Others: Conformity, Fads, and Informational Cascades, Journal of Economic Perspectives, Vol.12, No.3, pp.151 170 (1998). [3] DEIM Forum 2012, F8-2 (2012). [4] FIT2011 10 (2011). [5] Zhu, D., Fukazawa, Y., Karapetsas, E. and Ota, J.: Intuitive Topic Discovery by Incorporating Word-Pair s Connection into LDA, Proc. 2012 IEEE/WIC/AIM International Conferences on Web Intelligence and Intelligent Agent Technology, pp.303 310 (2012). [6] Vol.45, No.3, pp.919 933 (2004). [7] Sakai, H. and Masuyama, S.: Cause information extraction from financial articles concerning business performance, IEICE Trans. Inf. Syst., E91-D(4), pp.959 968 (2008). [8] Sakaji, H., Sakai, H. and Masuyama, S.: Automatic extraction of basis expressions that indicate economic trends, Proc. PAKDD 2008, pp.977 984 (2008). [9] 14 pp.1144 1147 (2008). [10] Adomavicius, G. and Tuzhilin, A.: Context-aware recommender systems, Recommender systems handbook, pp.217 256 (2011). [11] 2013 (2013). [12] 24 7 (2012). [13] Imamura, K., Kikui, G. and Yasuda, N.: Japanese dependency parsing using sequential labeling for semispoken language, Proc. ACL, pp.225 228 (2007). [14] (1977). 8. c 2015 Information Processing Society of Japan 25
2010 2012 NTT 1995 1997 ACM 1997 1999 2007 ACM 1989 c 2015 Information Processing Society of Japan 26