N-gram Language Models for Speech Recognition
|
|
|
- たけなり うみのなか
- 5 years ago
- Views:
Transcription
1 N-gram Language Models for Speech Recognition Yasutaka SHINDOH ver
2 N-gram 5. N-gram0 6. N-gram 7. 2-gram vs. 3-gram vs. 4-gram 8. 9.
3
4 (1) name twitter web site interest, favorite Mac OS X, GNU Emacs, GNU Make
5 (2) / P (ICT) Windows M () HMI K(, K)
6 (3) (, ) ü (non-commutative rings) 0 1 ü (modules) () 0 D3! 1!
7
8 (1) () MFC, PLP () () (decoder) HMM () (1 or 0)
9 (2) (MFC) (LPCM) () (MFC) FFT (Mel Freq. Filter) DCT
10 (3) (MFC) ()
11 (4) () ()
12
13 (1) ()!!?
14 (2) (CFG)!!?
15 (3) () (N-gram)!!!?
16 N-gram
17 N-gram (1) () () 2 (2-gram) 3 (3-gram)
18 N-gram (2) () jp 2009/07/05 (DB) () () ()
19 N-gram (3) () () count N-gram () () () N-gram ()
20 N-gram (4)!
21 N-gram0
22 N-gram0 (1) N-gram () (: ) (: ) N-gram () (: LDAPLSA) (: CRF) N-gram (: Class N-gram)
23 N-gram0 (2) N-gram 2-gram (0)! P( ) P( ) P( ) P( ) P( ) P( )
24 N-gram0 (3) 0 ( w 1 w 2 P(w 2 w 1 ) λ (0<λ<1) w 1 w 2 ΣP(w 2 w 1 ) 1-λ λ Kneser-Ney Modified Kneser-Ney Witten-Bell Good-Turing Natural (!)
25 N-gram0 (4) N-gram 2 a) (: Word Correct, Word Accuracy) b) (: Perplexity) Perplexity Perplexity () Perplexity () : <s>,,,,,,, </s> : P( <s>) P( ) P( ) P( ) P( ) P( ) P(</s> ) Perplexity: (P( <s>) P( ) P( ) P( ) P( ) P( ) P(</s> )) -1/7
26 N-gram
27 N-gram (1) blog Perplexity MeCab ver NAIST Japanese Dictionary ver SRILM ver
28 N-gram (2) ()
29 N-gram (3) () Wikipedia
30 N-gram (4) ( (1-gram)) V = KS a S: V: K, a: (0.4<a<0.6)
31 N-gram (5) (1-gram)
32 2-gram vs. 3-gram vs. 4-gram
33 2-gram vs. 3-gram vs. 4-gram (1) Kneser-Ney: / Perplexity gram 3- gram 4- gram
34 2-gram vs. 3-gram vs. 4-gram (2) Modified Kneser-Ney: / Perplexity gram 3- gram 4- gram
35 2-gram vs. 3-gram vs. 4-gram (3) Witten-Bell: / Perplexity gram 3- gram 4- gram
36 2-gram vs. 3-gram vs. 4-gram (4) Perplexity! 2-gram >>>>> 3-gram > 4-gram3-gram4-gram 3-gram Kneser-Ney, Modifiled Kneser-Ney, Witten-Bell
37
38 (1) Kneser-Ney: / 150! 130 Perplexity gram 3- gram 4- gram
39 (2) Kneser-Ney: / 150? 130 Perplexity gram 3- gram 4- gram
40 (3) Kneser-Ney: / 150? 130 Perplexity gram 3- gram 4- gram
41 (4) 1.5Peplexity 2Perplexity Perplexity Perplexity
42
43 : Julius ver : SRILM ver gram+4-gram Modified Kneser-Ney (1-gram) blog (+)
44
フリーソフトでつくる音声認識システム ( 第 2 版 ) サンプルページ この本の定価 判型などは, 以下の URL からご覧いただけます. このサンプルページの内容は, 第 2 版 1 刷発行時のものです.
フリーソフトでつくる音声認識システム ( 第 2 版 ) サンプルページ この本の定価 判型などは, 以下の URL からご覧いただけます. http://www.morikita.co.jp/books/mid/084712 このサンプルページの内容は, 第 2 版 1 刷発行時のものです. i 2007 10 1 Scilab 2 2017 2 1 2 1 ii 2 web 2007 9 iii
Twitter‡É‡¨‡¯‡éŁ”¦…V…X…e…•‡Ì™ñ‹Ä
Twitter 1 Twitter Twitter Twitter Twitter Streaming API 18 2 1 6 1.1.................................................. 6 1.2................................................. 7 1.3..................................................
受付案内ロボットASKA
ASKA Copyright 2002 NAIST 83.5 86.3 1.1 10.1 ASKA 42.3 52.8 1.6 51.2 88.0 89.9 0.3 26.5 93.7 94.7 4.0 50.5 % % % RWCP) NEDO 30 ASKA ASKA ASKA 2001.7 NHK 2001.10 Web JNAS(306) 40kwords PTM JULIUS
¥ì¥·¥Ô¤Î¸À¸ì½èÍý¤Î¸½¾õ
2013 8 18 Table of Contents = + 1. 2. 3. 4. 5. etc. 1. ( + + ( )) 2. :,,,,,, (MUC 1 ) 3. 4. (subj: person, i-obj: org. ) 1 Message Understanding Conference ( ) UGC 2 ( ) : : 2 User-Generated Content [
/27 (13 8/24) (9/27) (9/27) / / / /16 12
79 7 79 6 14 7/8 710 10 () 9 13 9/17 610 13 9/27 49 7 14 7/8 810 1 15 8/16 11 811 1 13 9/27 (13 8/24) (9/27) (9/27) 49 15 7/12 78 15 7/27 57 1 13 8/24 15 8/16 12 810 10 40 1 Wikipedia 13 8/18, 8/28 79
JA2008
A1 1 10 vs 3 2 1 3 2 0 3 2 10 2 0 0 2 1 0 3 A2 3 11 vs 0 4 4 0 0 0 0 0 3 6 0 1 4 x 11 A3 5 4 vs 5 6 5 1 0 0 3 0 4 6 0 0 1 0 4 5 A4 7 11 vs 2 8 8 2 0 0 0 0 2 7 2 7 0 2 x 11 A5 9 5 vs 3 10 9 4 0 1 0 0 5
( ) ver.2015_01 2
1 1.1 1.2 1.3 2 ( ) 2.1 2.2 2.3 2.4 3 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5 5.1 5.2 5.3 5.4 6 6.1 6.2 6.3 7 7.1 7.2 7.3 8 ver.2015_01 2 1 1.1 1.2 1.3 ver.2015_01 3 2 2.1 2.2 2.3 ver.2015_01 4 2.4 ver.2015_01
untitled
10 2 12!! 12!! 3 4 ( 5 6 Mapion http://www.mapion.co.jp/ 7 8 1.3km 9 30!? 10 130 160 Twitter 11 13 6 http://www.jiyugaoka.or.jp/ or jp/ 10 Wi-Fi 9 12 13 14 ( 40 2 1 2,000 30 4,000 60,, 105 1 15 http://www.nakanobu.com/
Microsoft Word - ランチョンプレゼンテーション詳細.doc
PS1-1-1 PS1-1-2 PS1-1-3 PS1-1-4 PS1-1-5 PS1-1-6 PS1-1-7 PS1-1-8 PS1-1-9 1 25 12:4514:18 25 12:4513:15 B PS1-1-10 PS1-2-1 PS1-2-2 PS1-2-3 PS1-2-4 PS1-2-5 PS1-2-6 25 13:1513:36 B PS1-2-7 PS1-3-1 PS1-3-2
MFC-100 Version E ... 1 2 3 4 1 2 3 1 1 1 2 2 3 3 1 1 2 2 3 3 1 2 3 1 2 3 1 2 3 4 1 1 1 2 3 2 3 1 1 3 2 1 1 2 1 1 2 3 4 1 2 3 4 5 6 1 2 6 3 4 5 1 2 3 1 2 4 1 2 3 1 2 3 1 2 5 3 6 7
QW-3414
MA1312-C P 1 2 3 A E L D E D A A E D A D D D D D E A C A C E D A A A C A C A C E E E D D D A C A C A A A A C A C A C E E C C E D D C C C E C E C C E C C C E D A C A C A C E L B B
untitled
ICT ICT 1 ICT 2 ICT 3 ICT 4 ICT 5 ICT 6 ICT 7 ICT 8 ICT 9 ICT 10 ICT 11 ICT 12 ICT 13 ICT 14 ICT 15 ICT 16 ICT 17 ICT 18 ICT 19 ICT 20 ICT 21 ICT 22 ICT 23 ICT 24 ICT 25 ICT 26 ICT 27 ICT 28 ICT 29 ICT
- 1 - - 0.5%5 10 10 5 10 1 5 1
- - - 1 - - 0.5%5 10 10 5 10 1 5 1 - 2 - - - - A B A A A B A B B A - 3 - - 100 100 100 - A) ( ) B) A) A B A B 110 A B 13 - 4 - A) 36 - - - 5 - - 1 - 6-1 - 7 - - 8 - Q.15 0% 10% 20% 30% 40% 50% 60% 70%
untitled
186 17 100160250 1 10.1 55 2 18.5 6.9 100 38 17 3.2 17 8.4 45 3.9 53 1.6 22 7.3 100 2.3 31 3.4 47 OR OR 3 1.20.76 63.4 2.16 4 38,937101,118 17 17 17 5 1,765 1,424 854 794 108 839 628 173 389 339 57 6 18613
untitled
1. 3 14 2. 1 12 9 7.1 3. 5 10 17 8 5500 4. 6 11 5. 1 12 101977 1 21 45.31982.9.4 79.71996 / 1997 89.21983 41.01902 6. 7 5 10 2004 30 16.8 37.5 3.3 2004 10.0 7.5 37.0 2004 8. 2 7 9. 6 11 46 37 25 55 10.
120218柔道死亡の新聞報道泰平作成
Ver.01.07 Last Upd:2011/12/03 1 2 3 4 5 1994 08 05 1994/08/05 31 414 6 1995 03 28 7 1987/06/23 8 8 1987 07 28 011 00419 1988/07/07 15 340 9 1989 05 18 023 00579 10 1989 07 17 00440 11 1989 10 31 00232
<4D6963726F736F667420576F7264202D2081A193B98BE257656290EA97708CFB8DC08B4B92E88179918D8D878CFB8DC0817A816990568B4B816A81798A6D92E894C5817A2E646F63>
夏目小兵衛直克
39(1906)1222 14(1817) 3(1832)1514(1843) 2628 6 (1853) (1854)3727 3(1856) 1 / 13 5(1858)6(1859) 5(1853) () () () () () () 3(1867)29 504111( 2 / 13 )98 23 18 2(1869)310283 100 50() 58 226 3313200982 5033
nenkin.PDF
1 31 1 WEB 10 3,544 429 13 10 22 11 7 WEB 1 2 41.0 15 80.0 20 46.7% 1000 55.8 1000 34.4 21 18.2 1000 23 25 41.0 49.2 29 90.6 42.7 33 56.4% 79.2% 67.4 51.7 37 39 83.7 1 91.0 93.6 9 2 3 1000 96.3 300 1000
-1-
-1- -2- -3-1 8 6% 2 4 6 8 1 48 63 43 6 55 38 78 58 2 88 67 11 22 78 1 56 22 89 47 34 36 32 38 4 34 26 7 -4- 18-5- 3 25 28 (6.%) (6.%) (.9%) 2 15 18 158 1 (3.8%) (56.4%) 5 2 137 27 8 1 68 119 26 71 28 65
( )
Web Web 1 3 1 21 11 22 23 24 3 2 3 4 5 1 1 11 22 9 2 3 15 11 22 2 11 21 4 5 ( ) 102 ( ) 1 ( 1 2001 Web 1 5 4 1 1 - 7 - [] - 7 10 11 12 12 1 10 1 12 - [] 1 1 2 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 3 1 47
fb_town
Welcome to Facebook Town! Facebook 0 ver.2 http://www.facebook.com/tsukasa.kawashima 20 mac os0.6.6 apple pages epub 2.0. pages Microsoft Windows Microsoft Corporation Apple Apple itunes iphone ipad Apple,Inc.
tonan-cs.indd
J I S adobe japan 1-4 00000000000000000000 000 1986 1927 1999 1996 2001 2007 2003 2006 2003 2007 19 1956 1996 24 1996 2000 1991 1979 40 2000 2006 1996 2001 1978 2003 2006 2000 2002 2004 2006
Microsoft Word - SNSで繋がる人間関係.doc
SNS SNS ID SNS ixi Twitter Facebook SNS SNS SNS SNS SNS SNS Mixi Mixi Mixi SNS SNS SNS Mixi 2009 9 30 1,792 SNS 52.2% 47.8% 20 24 33.8% 25 29 28.4% 30 34 17.6% Wikipedia SNS Mixi Mixi SNS Mixi SNS Twitter
2 Part A B C A > B > C (0) 90, 69, 61, 68, 6, 77, 75, 20, 41, 34 (1) 8, 56, 16, 50, 43, 66, 44, 77, 55, 48 (2) 92, 74, 56, 81, 84, 86, 1, 27,
/ (1) (2) (3) ysawano@tmu.ac.jp (4) (0) (10) 11 (10) (a) (b) (c) (5) - - 11160939-11160939- - 1 2 Part 1. 1. 1. A B C A > B > C (0) 90, 69, 61, 68, 6, 77, 75, 20, 41, 34 (1) 8, 56, 16, 50, 43, 66, 44,
PageScope Box Operator Ver. 3.2 Box Operator !. - - 2! - - 2 - 2 - - - - - - - - - - - - - 2 2-2 2-2 - - - 1 2 3 4 2 - 2 - - - - - - - - - - 2 - - - - - - - - - 2 0 - - 2 0 - - 2 0 - -
01_表紙
2012.1.10 8 Hijicho http://hijicho.comhttp://twitter.com/hijicho/ Hijicho http://hijicho.comhttp://twitter.com/hijicho/ () () Hijicho http://hijicho.comhttp://twitter.com/hijicho/ ( ) Hijicho http://hijicho.comhttp://twitter.com/hijicho/
gengo.dvi
4 97.52% tri-gram 92.76% 98.49% : Japanese word segmentation by Adaboost using the decision list as the weak learner Hiroyuki Shinnou In this paper, we propose the new method of Japanese word segmentation
IPSJ SIG Technical Report Vol.2013-SLP-98 No /10/25 1,a) 1 ( Q&A ) ( ) YJVOICE Development of speech recognition and natural language processing
1,a) 1 ( Q&A ) ( ) YJVOICE Development of speech recognition and natural language processing for ONSEI Assist service Abstract: ONSEI Assist is a voice dialog application for mobile devices that enables
¥ƥ¥¹¥ȥ¨¥ǥ£¥¿¤λȤ¤˽
: 2010 2 14 1 MS Word.doc (MS Word 2003 ).docx (MS Word 2007 ) Word Windows.txt MS Word Word Word Word Excel Word 1 Word Word Word MS Word MS Word MS Word Word Windows MS Word MS Word Word Windows.txt
C-DSD48.dvi
21 Twitter Candidate Utterance Acquisition Method for Non-task-oriented Dialogue Systems from Twitter Michimasa Inaba Sayaka Kamizono Kenichi Takahashi Hiroshima City University inaba@hiroshima-cu.ac.jp
FileMaker Pro Tutorial
FileMaker Pro 12 2007-2012 FileMaker, Inc. All Rights Reserved. FileMaker, Inc. 5201 Patrick Henry Drive Santa Clara, California 95054 FileMaker Bento FileMaker, Inc. Bento FileMaker, Inc. FileMaker FileMaker,
untitled
FAIR 2007 4 FAIR 2007 4 FAIR Web / FAIR FAIR Finance and Accounting Information Rules FAIR 2007 FAIR FAIR FAIR FAIR 2007 4 1. FAIR 1.1 FAIRWeb 1.2 1.3 FAIRWeb 1 1 2 2. 2.1 2.2 2.3 2.4 2.5 2.5.1 2.5.2 2.5.3
( : A9TB2096)
2012 2013 3 31 ( : A9TB2096) Twitter i 1 1 1.1........................................... 1 1.2........................................... 1 2 4 2.1................................ 4 2.2...............................
21 Pitman-Yor Pitman- Yor [7] n -gram W w n-gram G Pitman-Yor P Y (d, θ, G 0 ) (1) G P Y (d, θ, G 0 ) (1) Pitman-Yor d, θ, G 0 d 0 d 1 θ Pitman-Yor G
![21 Pitman-Yor Pitman- Yor [7] n -gram W w n-gram G Pitman-Yor P Y (d, θ, G 0 ) (1) G P Y (d, θ, G 0 ) (1) Pitman-Yor d, θ, G 0 d 0 d 1 θ Pitman-Yor G](/thumbs/89/99361107.jpg "21 Pitman-Yor Pitman- Yor [7] n -gram W w n-gram G Pitman-Yor P Y (d, θ, G 0 ) (1) G P Y (d, θ, G 0 ) (1) Pitman-Yor d, θ, G 0 d 0 d 1 θ Pitman-Yor G")
ol2013-nl-214 No6 1,a) 2,b) n-gram 1 M [1] (TG: Tree ubstitution Grammar) [2], [3] TG TG 1 2 a) ohno@ilabdoshishaacjp b) khatano@maildoshishaacjp [4], [5] [6] 2 Pitman-Yor 3 Pitman-Yor 1 21 Pitman-Yor
Microsoft Word - cookpad2006
SIV 2006 12 COOKPADhttp://cookpad.com/1998 3 16 (PV) 1 284 PV/ 167 2006 11 No.1 COOKPAD COOKPAD WEB WEB COOKPAD COOKPAD COOKPAD COOKPAD 1 COOKPAD WEB COOKPAD 2 WEB2.0 COOKPAD Web Web2.0Web2.0 Web Web 2001
IPSJ SIG Technical Report Vol.2010-NL-199 No /11/ treebank ( ) KWIC /MeCab / Morphological and Dependency Structure Annotated Corp
1. 1 1 1 2 treebank ( ) KWIC /MeCab / Morphological and Dependency Structure Annotated Corpus Management Tool: ChaKi Yuji Matsumoto, 1 Masayuki Asahara, 1 Masakazu Iwatate 1 and Toshio Morita 2 This paper
KLCシリーズ インストール/セットアップ・ガイド
KORG Legacy Collection J 1 / 2 Windows XP Windows XP 1 2 Windows XP 3 4 5 6 3 / 7 8 4 Mac OS X Mac OS X 1 2 3 4 5 Mac OS X 6 5 / 7 8 6 USB / USB 1 2 USB 7 / 1 2 3 Windows Mac 4 1 Windows Mac 8 2 Windows
Windows 10 2 WaaS IT Windows 10 WaaS Windows as a Service Windows 10 WaaS IT IT IT IT 2
Windows 10 2 WaaS IT Windows 10 WaaS Windows as a ServiceWindows 10 WaaS IT IT IT IT 2 3 4 WaaS Windows 10 1. OS 2. Windows 10 3. OS UX UWP BLOG Part 2. Windows 10 https://blogs.msdn.microsoft.com/nakama/2016/08/18/win10waas-part2/
( ) Kevin Duh
NAIST-IS-MT1251045 Factored Translation Models 2014 2 6 ( ) Kevin Duh Factored Translation Models Factored translation models Factored Translation Models, NAIST-IS-MT1251045, 2014 2 6. i Post-ordering
Microsoft Word - H22_Info_Pro_Pre_Enquete_01.docx
22(2010) 2010 4 26 PC 106 185 C11 78 C13 155 C21 221 C13 172 9 65 C21 237 C21 187 C13 140 1,449 1 180 169 2 422 235 413 1,419 Q PC Q Q Q PC OS Windows Q Mac OS X Q Linux Q Q Q Windows OS Windows 7 Windows
Vol. 43 No. 7 July 2002 ATR-MATRIX,,, ATR ITL ATR-MATRIX ATR-MATRIX 90% ATR-MATRIX Development and Evaluation of ATR-MATRIX Speech Translation System
Vol. 43 No. 7 July 2002 ATR-MATRIX,,, ATR ITL ATR-MATRIX ATR-MATRIX 90% ATR-MATRIX Development and Evaluation of ATR-MATRIX Speech Translation System Fumiaki Sugaya,,, Toshiyuki Takezawa, Eiichiro Sumita,
orpheewe決定版
b b b b b b 1 Allegro con uoco Allegro con uoco 2 9 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 17 4 25 5 31 6 Allegretto b 9 Allegretto 7 38 rit. Lento rit. Lento 8 49 9 59 10 66 11 Allegro Vivace 3 3 3
0_テキストマイニング環境構築_mac
本実験で利 するソフトウェア Anaconda 5.0.0 Python 本体及び数値計算ライブラリや開発環境 をセットにしたパッケージ MeCab IPA 辞書 NAIST Japanese Dictionary mecab-ipadic-neologd Natto-py 日本語形態素解析器 Mecab 用辞書 Mecab 用辞書 Mecab 用辞書 Python から MeCab を利用するためのライブラ
00 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.... 0........ 0 0 0 0 0 0 0 0 0 0..0..........0 0 0 0 0 0 0 0 0 0 0.... 0........ 0 0 0 0 0 0 0 0 0 0... 0...... 0... 0 0 0 0 0 0..0 0... 0 0 0 0 0.0.....0.
Emacs ML let start ::= exp (1) exp ::= (2) fn id exp (3) ::= (4) (5) ::= id (6) const (7) (exp) (8) let val id = exp in
Emacs, {l06050,sasano}@sic.shibaura-it.ac.jp Eclipse Visual Studio Standard ML Haskell Emacs 1 Eclipse Visual Studio variable not found LR(1) let Emacs Emacs Emacs Java Emacs JDEE [3] JDEE Emacs Java 2
ユーザ負担のない話者・環境適応性を実現する自然な音声対話処理技術
DSP DSP Julius H18 H16 H19 / (2003.3) (2004.10) Julian(2005.08) JNAS,.,. CSRC,..,.. 2006 8 1 8 20 1 1166 451 56 54 605 /2,800msecshort reject 2006.3.27 Julius #Downloads per month 3.5 3.5 3.4 3.5 3.5
( ) ID -S (Conflict Of Interest, COI) a-d COI a. b ( ) ( ) c d
2017 Abstract 2017 9 5 6 ( ) 2017 2017 9 7 ( ) 2017 9 5 6 ( ) 2017 https://goo.gl/45fhez ID -S 10 1 9 5 9 6 (Conflict Of Interest, COI) a-d COI a. b. 2014 10 ( ) ( ) c. 2014 10 d. 2014 10 2 Programme:
大阪府私学教育情報化研究会
2003 ICT http://www.osaka-sigaku.net 1. 209 209 0. 529,000 529,000 0. 0 0 0. 0 0 0. 1 0 1 529,210 529,209 1. 175,000 200,000-25,000. 97,000 50,000 47,000. 8,645 12,000-3,355. 0 20,000-20,000. 94,344 70,000
IPSJ SIG Technical Report Pitman-Yor 1 1 Pitman-Yor n-gram A proposal of the melody generation method using hierarchical pitman-yor language model Aki
Pitman-Yor Pitman-Yor n-gram A proposal of the melody generation method using hierarchical pitman-yor language model Akira Shirai and Tadahiro Taniguchi Although a lot of melody generation method has been
<834A8358835E837D83438359837D836A83858341838B2D30312E6169>
01 1 2 3 4 02 03 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 04 1 2 1 2 3 3 05 06 01 02 03 04 01 02 03 04 05 SITE Web Site Welcome to my home page. SITE Web Site Welcome to my home page.
音響モデル triphone 入力音声 音声分析 デコーダ 言語モデル N-gram bigram HMM の状態確率として利用 出力層 triphone: 3003 ノード リスコア trigram 隠れ層 2048 ノード X7 層 1 Structure of recognition syst
1,a) 1 1 1 deep neural netowrk(dnn) (HMM) () GMM-HMM 2 3 (CSJ) 1. DNN [6]. GPGPU HMM DNN HMM () [7]. [8] [1][2][3] GMM-HMM Gaussian mixture HMM(GMM- HMM) MAP MLLR [4] [3] DNN 1 1 triphone bigram [5]. 2
Microsoft Word - 倫理 第40,43,45,46講 テキスト.docx
6 538 ( 552 ) (1) () (2) () ( )( ) 1 vs () (1) (2) () () () ) ()() (3) () ( () 2 () () () ()( ) () (7) (8) () 3 4 5 abc b c 6 a (a) b b ()() 7 c (c) ()() 8 9 10 () 1 ()()() 2 () 3 1 1052 1051 () 1053 11
