Vol. 52 No. 3 1033 1044 (Mar. 2011) IME 1 2 1 1 IME Web PC Android Dynamic Dictionary Generation Method for Context-aware Input Method Editor Yutaka Arakawa, 1 Shinji Suematsu, 2 Shigeaki Tagashira 1 and Akira Fukuda 1 In this paper, we propose a concept of context-aware IME (Input Method Editor) for improving the input of Japanese on mobile devices. In our concept, according to the user s current location, a personal context-aware dictionary is dynamically generated from the keywords gotten via some APIs in the Internet. We propose dynamic dictionary generation method, system architecture, sorting algorithm based on a result of web search. The result of response time of our proto type systems show its effectiveness. 1 Graduate School of Information Science and Electrical Engineering, Kyushu University 2 Faculty of Information Science and Electrical Engineering, Kyushu University 1. iphone Android PC PC 1) 5 8 Google 1 2 3 4 1 1 T9 2) 2 Social IME 3) 3 PObox 4) 4 IME iwnn 5) 1033 c 2011 Information Processing Society of Japan
1034 IME Web API Web API 1 IME Fig. 1 Suitable situations for context-aware IME. IME IME IME IME 1 GPS Web API Web API Web API Web 2 4 5 6 7 2. 2.1 6) 7) web
1035 IME web iwnn 8) Kukura 9) Kukura POBox 2.2 MS-IME ATOK Google IME Baidu Type Web 3) Social IME Google IME Social IME 2.3 10) RFID 11) 12) 13) 14),15) 16) Synapse 17) 18) 2.4 Web 1 Web Web API Web API SOAP REST XML Web API 19) 3. Twitter 20) 21) Twitter 2006 7
1036 IME Fig. 3 3 Geographical distribution of tweets including Shibuya and Shinjuku. 2 Fig. 2 How to analyze the information obtainable from tweets. Obvious 140 GPS Geo API 2 1 Yahoo! API Yahoo! API 4.83% 13,590 656 20 1 3 4. 1) 2) 3) Web
1037 IME (2) Web API (3) (4) (5) Web Web API Web 4 Yahoo! API 22) Google Maps API 23) API 24) Web API MeCab 4.2 4 Fig. 4 Construction of our proposed system. 4.1 4 IME Web GPS IME (1)
1038 IME API 4.3 6) Web API Social IME 3) Web Web API Web API Yahoo! API 22) Product Advertising API Amazon 25) API 24) Google Maps API 23) Mecab 26) 4.4 Yahoo! API Yahoo! API Yahoo! API Twitter X Y X Z 5. 4 IME 5.1 2 GPS Google Maps API PC HTML IME 2 ATOK 27) PC ATOK ATOK Windows Mac OS 5 IME ATOK 7 ATOK ATOK 6 Android OpenWnn Android HT-03A Android SDK 1.6 HTC Hero Android SDK 2.0 OpenWnn IME IME
1039 IME 5 ATOK Fig. 5 Screenshot of our prototype working as ATOK Direct Plugin. Fig. 7 7 IME Screenshot of status of the server for context-aware IME. ruby1.8 Ubuntu9.06 SQLite MeCab 26) Web API Yahoo! 6 Android OpenWnn Fig. 6 Screenshot of our extended OpenWnn on Android. 5.2 IME IME Web API API 7 IME Google Map Yahoo! API Yahoo! API
1040 IME 6. Web 6.1 API IME 1 sec/letter 28) 1) API 2) 3) Yahoo! API 2) IME MeCab 26) T API T d T r T t T d 266 T r T t 23 (a) (i) (u) (e) (o) T t E-mobile T d 1 8 Yahoo! API MeCab 1 Table 1 Result of the size of generated dictionary and the time for generation T d. 38 503 238.8721805 0.487469 3.019088 1.860482139 Fig. 8 8 Distribution of dictionary generation time for the size. 1 38 503 238.9 0.5 3 1.9 Yahoo! API 1 100 MeCab MeCab NAIST JDIC 0.6.2-20100208
1041 IME Table 2 2 T T r T t Result of T, T r and T t for the number of candidate words. 0 24 6.026785714 T 0.469193 1 15.26841 18 0.848508813 T r 0.186305 19 0.264816 18 0.192301478 T t 0.277089 5 15.056005 5 0.656213214 Fig. 10 10 Examples of evaluation sheet. 2 9 2 0 24 6 9 T T r T t Fig. 9 Distribution of T, T r and T t for the number of candidate words. T d 8 266 1 0.5 15 T r 0.19 0.26 T t E-mobile 9 T T r T t T r T T t
1042 IME 11 12 Yahoo! Fig. 11 The number of useful words by users. Fig. 12 Cumulative distribution of the rank of corresponding words sorted by Yahoo! Search. 6.2 Web 20 10 2 API Yahoo! API 20 1 1 10 400 4 11 1.7 Yahoo! API Twitter 1 Yahoo! API 12 10 45%
1043 IME web 7. IME PC Android Twitter Google Map Yahoo! API MeCab 0.85 sec Web 1) rtype PC 5 rtype (2009). http://release.center.jp/2008/11/0502.html 2) Grover, D., King, M. and Kuschler, C.: Patent No.US5818437, Reduced keyboard disambiguating computer, Tegic Communications, Inc., Seattle, WA (1998). 3) 190 (2009). 4) Masui, T.: POBox: An Efficient Text Input Method for Handheld and Ubiquitous Computers, Lecture Notes in Computer Science, Vol.1707/1999, pp.289 300, Springer Berlin / Heidelberg (1999). 5) iwnn. http://www.omronsoft.co.jp/sp/ 6) 17 DEWS2006 pp.2b i10 (2006). 7) Web (2007). 8) Vol.44, No.11, pp.2538 2546 (2003). 9) Kukura WISS2002 pp.43 47 (2002). 10) RFID DBSJ Vol.6, No.3, pp.13 16 (2007). 11) 94 (2008). 12) IN Vol.104, No.691, pp.149 154 (2005). 13) DB DBSJ Vol.7, No.1, pp.221 226 (2008). 14) 2 Vol.2006, No.2, p.264 (2006). 15) B Vol.88, No.11, pp.2137 2146 (2005). 16) Synapse 2 Vol.2005, No.2, p.537 (2005). 17) Vol.75, No.1, pp.94 97 (2008). 18) NTT DoCoMo Vol.17, No.3, pp.55 61 (2009-10). 19) Nishio, N., Sakamoto, N. and Arai, I.: Real-World-oriented Contents Management System Mashed up with Google Street View, Demonstration at Pervasive (2009). 20) Twitter Twitter. http://twitter.com/ 21) Twitter Vol.2010-MBL-53, No.50, pp.1 7 (2010). 22) Yahoo! (2009). http://developer.yahoo.co.jp/webapi/map/ 23) Google Code (2009). http://code.google.com/intl/ja/apis/ maps/ 24) Web (2009). http://api.gnavi.co.jp/api/
1044 IME service.htm 25) Product Advertising API (2009). https://affiliate.amazon.co.jp/gp/advertising/api/sdetail/main.html 26) MeCab: Yet Another Part-of-Speech and Morphological Analyzer (2009). http://mecab.sourceforge.net/ 27) Corporation, J.: ATOK.com ATOK (2009). http://www.atok.com/ 28) Vol.8, pp.23 32 (2006). ( 22 5 21 ) ( 22 10 4 ) 1977 2001 2003 2004 COE 2006 2006 2007 2009 3 2010 4 LSI APCC 2008 Best Paper Award 2008 MBL 2009 DICOMO 2010 DICOMO 2010 IEEE 1985 2008 2010 1996 1998 2000 2000 2007 2009 2009 IEEE 1977 1979 NTT 1983 1989 1994 2001 2008 LSI 1990 Best Author 1993 ACM IEEE Computer Society OR