ࡱ>  ,bjbjΚΚ 48?$H||> & & & < !g",".>0>0>0>0>0>0>,@C\>"!"!""\>#||& < wq>###"v|8& < .>#".>##6;=_!#6<>>0><D#D8=#= """\>\>#""">D""""""""" B:   Searching Past Search Engines By Steven Bethard Institute of Cognitive Science and Department of Computer Science It wasnt so long ago that if you wanted to, say, find out the population of Boulder, youd go to your local library, find a book with census and demographic data, and then page through until you found the city and figures you were interested in. But with the advent of the internet and search engines, bits of knowledge like these are now just a few keywords away. Typing population Boulder into Google gives back a results page where the first click lands you at the Boulder County Census page, telling you that in 2000, there were exactly 94,673 people in Boulder. Web search engines answer questions like these with such ease that were often surprised when a search engine makes us work a bit to get our answer. Say we want to know: Were the World Trade Center towers standing when the plane crashed in Pennsylvania? We put the question to Google, and it gives us back thousands of pages about the September 11 attacks. On the one hand thats good, since were pretty sure one of those pages has our answer in it. But commercial search engines like Google or Yahoo dont really understand what were looking for, so its up to us to go through those pages onebyone until we find the answer to our question. In this particular case, it takes about two pages of text to find out that the south tower fell at 9:59am, UA 93 crashed in Pennsylvania at 10:03:11, and the north tower fell at 10:28am. Many researchers in natural language processing (NLP) are interested in making web systems that can answer difficult questions like these as quickly and easily as Google currently answers factoid questions. This is a difficult task because computers really know very little about how language works. Consider text like: The top commander of a Cambodian resistance force said Thursday he has sent a team to recover the remains of a British mine removal expert kidnapped and presumed killed by Khmer Rouge guerrillas almost two years ago. Most search engines represent sentences like these as a simple bag of words which says only that the word a appeared three times, the word of, two times, the word Cambodian, once, etc. Such a simple representation is often enough to answer basic factoid questions, but falls short when faced with complex questions that require an understanding of events and the order in which they occurred. Recent NLP research has focused on a number of ways to move past this simple word-level understanding of text. Rather than treating documents like bags of unordered words, these researchers look at how computers can understand how language works to identify entities, events and the relations between them. So the sentence above would be represented as a graph (see  REF _Ref165729095 \h Figure 1), where we can see relations like: The victim of the kidnapped was the expert The commander sent his team out before he said anything about it Because these graphs include much more detailed information about how the different words in a document are related to each other, systems based on them can answer much more sophisticated questions than the simple bag of words approach can. However, extracting such graphs from text also requires much more sophisticated techniques. Where the bag of words approach needed only some basic knowledge of how spacing and punctuation work, graph-based techniques require knowledge of how human language works. To know that things like theWorldTradeCenter should be treated as a single entity rather than four individual words, graph-based systems must be able to tell which words are people, locations and events and which are not. Researchers in the Institute of Cognitive Science (ICS) have shown recently that such semantic information can be extracted automatically using machine learning methods [1, 2, 3]. They start with large-scale document collections where humans have manually identified each person, location and event mentioned in the texts. These annotated documents, along with simple information such as which words are capitalized and which words are nouns or verbs, are then fed to machine learning algorithms which look for patterns in the data. These patterns are assembled into models which know enough about how language works that they can look at a new document and find the entities and events within it. Once the entities and events in a document are identified, the real work begins. To really be able to answer complex questions about what the document means, it is crucial that systems identify not just the entities and events, but the relations between them. Work on identifying relations is still very much in its infancy, and state-of-the-art systems can typically handle only a few different types of relations. One type of relation in which ICS researchers have made substantial progress is the predicate-argument relation [4]. These relations match events with the entities they concern. For example, in the sentences below, the verb hit is associated with an impactor relation to American Airlines Flight 11 and an impactee relation to the North Tower. American Airlines Flight 11 hit the North Tower. The North Tower was hit by American Airlines Flight 11. To identify such relations, computers must have a deep understanding of language structure. They must be able to realize that words may appear in different orders and still express the same basic meaning. The machine learning methods that learn such patterns no longer rely on simple features like which words are nouns or verbs. Instead, they look at things like how far the event and entity are from each other, and what kinds of words and phrases intervene. Only by delving a little deeper into how language works are these systems able to recognize entity-event relations. Of course, not every relation links an entity with an event. Of particular interest are temporal relations like before and after which relate a pair of events together. These are the kind of relations that allow systems to answer questions such as Which tower was hit first? or What order did the planes crash in? Again, ICS researchers have looked to machine learning techniques to address such relations. But temporal relations are harder than predicate-argument relations because they may relate events that are many phrases or even many sentences apart. Still, we know that humans have little trouble understanding such long distance relations, and in reading sentences like the one below, immediately recognize that the rising of share prices came after the resignation of Mr. Stronach: On the Toronto Stock Exchange yesterday, Magna shares rose 37.5 Canadian cents to C$9.625. Mr. Stronach, founder and controlling shareholder of Magna, resigned as chief executive officer last year. To be able to identify such temporal relations, ICS researchers are trying to use machine learning approaches that more closely follow the approach humans are believed to take. Long distance relations like the one between rose and resigned are built up by first identifying smaller relations, for example, that the rising was during the time yesterday and that the resignation was During the time last year. The system can then conclude that rising is after the resignation by simply recognizing that the time yesterday is after the time last year. In order to build systems that can identify even the simplest temporal relations, researchers need to dig deep into how language works and give the machine learning algorithms a variety of information, including the words and phrases that intervene between the events, as well as whether each event already occurred, is occurring now, or may occur in the future. Of course, extracting this kind of information is now many levels past the simple space and punctuation analysis that search engines perform. Still, recent results are promising and suggest that machines can learn such relations, in some cases finding as many as 9 out of 10 temporal relations correctly. So how long will you have to wait until you can go to the web, type in Which WTC tower was standing when the plane crashed in Pennsylvania? and have the search engine spit back the simple phrase The north one? Such a thing is still many years away. But the basic infrastructure on which these future systems will depend is being built today. Machine learning techniques are teaching computers the intricacies of human language, and when these computers are grown up theyll be able to converse with humans in new, useful and meaningful ways. Bibliography 1. Detection of Entity Mentions Occurring in English and Chinese Text. Hacioglu, Kadri, Douglas, Benjamin and Chen, Ying. Vancouver, B.C., Canada: s.n., 2005. HLT/EMNLP-2005. 2. Automatic Time Expression Labeling for English and Chinese Text. Hacioglu, Kadri, Chen, Ying and Douglas, Benjamin. Mexico City, Mexico: s.n., 2005. CICLing-2005. 3. Identification of event mentions and their semantic class. Bethard, Steven and Martin, James H. 2006. Empirical Methods in Natural Language Processing. 4. Support Vector Learning for Semantic Argument Classification. Pradhan, Sameer, et al. 1, 2005, Machine Learning, Vol. 60, pp. 11-39.  Figure  SEQ Figure \* ARABIC 1: An entity-event relation graph. qrV -%&'KLPQV^gpvw{<=>?@ƽ~w h:7mh]Qh]Sh]Q6 hWh]Q#hh]Q5B*CJ\aJphOhh]Q6 h]Q6 jhXd*h]QUmHnHuh]QmHnHu hOh]Qjh]QUjh]QU h$(h]Q h5h]Q hh]Q hHh]Qh]Q.0r V -KwLFw ""$ $h^hgd]Q $^gd]Q $ & Fgd]Qh^hgd]Qgd]Qgd]Qgd]Q@ABCBI]^_` (5DEFGOP_bepuv`afklp!!!!!! " "B"F""""######$$$ú hljh]Qhljh]Q6hW\h]Q_HmH sH h]Q_Hhljh]Q_H hdUh]Qho $h]Q6 hh]Q hSh]Q h/h]Q h]Q6h]QmHnHu h:7mh]Qh]Q<$$)$2$@$K$P$V$`$i$$$$$$$$$$$$$$' ()) *>*x***+Z+~+++,?,@,A,H,I,_,`,a,b,,,,,޳h]QB*mHnHphujhOh]QB*UphhOh]QB*ph hOh]Q$j}hih]QU_HmHnHuh]Q5\mHnHuh]Q6]mHnHuh]QmHnHu hh]Qhz>%h]Q6h]Q h]Q62$'))u*++?,A,,,,,,$a$gd]Q $$a$gd]Qgd]Q $^gd]Q 21h:p]Q/ =!"#$% }DyK _Ref165729095{DdB5FFj  FA?Picture 4"GkcOwvxB9KKisB$az8댞~7Uii~>=(:at-חg?d ƴ]S/mkgtvVzdbkûo PǴe.)o}p><<5lx={n4{k_xu=*4l/]#ܵHHS.SmYT7' {ldILLUd߷;zyH:E>zx!r:?J:7Yi6_뺉X@ٛ([I+9PO ?H>ugxNw)q0tf/gZd)5ͣA|i&&p L$Ȧ_aThO_6M[]߁FvY%BԡΪx.u vOj7U5VkP`Y u5a x)Gq}V487](z^e VG;L`VF- u3h ~Cθ?kƤCW ZgMۯ:D=:vNLD`cXcͧs:͘s~5̷gn+ jCAz㓠BkV<5K'ạQ! %a AP-c}4އg< 6s<ԛ:ܸP ߙO^eoXcك O/ZFn1/Ԩz\U;9>3e]h򺖜LR;1(:UNTkYX3fP|w_3ݪVvgzFo fgQ]{PxXcxd`־ZF/s1̚ Mm/O`^q -# {q`6>_qթ8-fP4&mVPHȕ?y(N.)P ٔ+8kWU.dr `d Tp1Y-&Gpi M>z;Eu H8pE819WǙ Kps#5_ariTX,NrdS_b3[մ.29`1Fb9}DVɿ۲聦.\ Kha?a`JZ$6cKՕ v16CR$igWMv.4Uw*֧"/UgeXQ‡Vi~2(0 [Av;hǃ0"ږ3f=! k\seo Qs" ۅΔ%/t+坿KC6T >#OIGX*-C;gYS0!!$ZV2Ĝ1RojɗHDߣplkS E#~TQ8@c1b ͏N$T
P& 0dͧ>=!`{/n FfnP.s,LaOƻ9k/B)AKƀ""͝Bd[ΠՄRgfO$ Aa<<ɝ|7]C v:ݹ2uM{o@W77:d 1&UT5k͑Bkx6kݴ{&ko+1ǙgtK3&`,Z!ݴ^s#_㜊l]RR8ch>)f˟%6fC*wRTOu44aS;dP]F2B6QaRD،/rHcVm6Y\fc)NQx\ 2o39 ( G*q ;r?Ol`Im,eYwP;wcMǽtͽUK .ɛ&rB_7CHӵm&sı+?#+y[W">[ ;A5Lt5N&CM#NM&.Z~\Z{?zz*ەS<>LpU Td.0es08?kF' : - YsR6B.oHC-|2M&/\]qW??҇]=[E@ib2P}Ȳ{2Qmm #~%:S¿4soKqQF҇cQ^!K0:$\ @jԚIkR!-6-O@y0ږ.í6|ZA\ku+4w\`ldYE9<$,>t[~=+KB G>c]OڇlRS\i-*nmO+{AM$LUKBO/rQ%p1EEُdoէaNsINhic%J□ R^⧨mմ.%z% gnqgvw&NaC @o8 M/N^ 8aUvp2:-aLi;۶Ȥ,RBr63ncgQ;D܎F;p+^g 6bѺ<2!{I jaQ^fi77S1~N&MG6(雄ʬE"e/zW7`5>! 9^:MaFk(SKK13;m[fZTS][Z= :L/se?IDyoAs8ICh(wfNҕi,ͻOz-e!= -1 Q*9n6Ssb,I 7*#0 qFlW/lR[2&PR֐Oơڄ :Z{ | lhN+Ap 3% MAiO L JpDFDF/ Y?UPYkˠ_=nU6DE.`z8vp8T0Քf%4+PQ?EF4ߦY;ƈ7olK=NˡDw(gƛrxVt6aL Ֆl}?JaTɰ&$<G%|UW(Gӷ@kȚW+ϩ` gh= $dQ@9w$`ke}twNqgqv~؎<|@GE3BޑTc(ɴҖ%Ӥnh"QF0^`=&9 )6EO9kb0:hqR ~alLO;HΉ޷`viz'\y( # cP12Hjҗc,k9o] ϰbP.LөX7Mci(5pQG(嚾QgGIO9D(PZc}78?O@C0a`WWg;K靡0arkYX>7xOv(EHmɓlFN(tHǖ$#\ifvINߢit5Qh Y;ɬ ^>YBZ|sN@G(N Jl|d~I;k?rr(0zd!a9Dj۳i!ųcKnag{aGU*vIuy2ue8KI#(&xK]~2%&l=++Y=Nz4 }yl#ʪo4nlP`. )%̾zCVڢeH&Y辰h߀qn>~8Tjgqs0gຟ3@yŰGV99)E%*5nb{݊E^x̶e([޶mT${2ڢfSS#>yp>UCg|C':*9f3`b1D'gҨ\@Sl23ibZT!im @.tĉIWC+%i=6Zu#<"KhHYf ,c.,+A]Tqt_U,p9M(xO{Nۜ[b{=> B2 3 PxK}'bba;?=Do3O(w13U5[CkثO qn~2cٕbӣᮞF&`#0 Ֆl!Ay /n -rй/]g~)uV!M wQ+5@aW#u/ Ws_E/5yLHVbld4A̧mM6apKf{IΥT3DQ/z=B+Sͬ*hH2v0m!g}}ʡj !~=s.u-,Z S%`{sәík!7vU6wze^`\8OM,3jIVƛp ^¨/7>|=y9,hсL,Tݳ [>I ZNh:t&(pYPbb5bhƭ]MnYTu3N c9MQ7w#~) 4O6L?pu8!UB u08]z?ZZpFvTuTntySݪo (aЃ?f DhN):>};'M:IKlfu!uVϏWz%uZ>Rwm '| L@#|;kD Ϣ?DgvKk3ړ#vdC8>_YJ|3\G \87hzM3쭕+ $2u4mtHS ?3ԀsQF Z`k4:йNi4"O3:@uȼI"1wKe /^R3vELq?3FTg0S"I{1VQpCٯ:{y1Q6JCTs@RYD;3{$k2%?KS3*ҝ[tsdyI`Sf&U=ŀƾk1H[J^<,yŵL`'>Ή.)}ɺ4+!ڝ.ٳюSLۅ~@RyjR." ~R 6e-3v9k7AIgq8Km#=ܳ*ZX.El%P\ު+<~:ހsAaNZ$+LwX364z~]+s_-/؞1#(,D: R{9H U?o%Z4"!\1i,|l06]t( G0.mW;IFxqF';OY#)ȝx|{DJuT]:"Ӑ0FpDy#[)O rYöw%Zsxݞwkju|$zgx6œr8v S0){:ݬCbehi?0*v]=D,f0'^cNr.T#g\ҕo{$V"@D5~C8ąϼghH4C9[=ʉ. Yug0pg]O]D3F(MBֿjd>_NKObTN-J»iߴY8ՔN\%E+%;;RS} ێxEqQYn$qk5ԁ5C:蛀v| tCG~頯I1.1 M=d#i7I@%?+Y[4r>*7iN10׫Vbs#"l aQPI }zs9(m rE^Įlf𒸽q-WQ7og{XrWj!(+ME׽umlEflӔO~*ǍO7(R~ٝouTM$2 M*1S`aeUoG2gKZNv"#`Rz q z@%LrGeICL˘}[ r&Fd\$(3̎gm@*YUǾs숖ʨdZk:j|J-z}gG6_VBmMsM"`;<@JDYgy?k[7GHaNt2z DIb( G ;@CL2ush,& c1٦'W*9.ZO4yL:U|ySV[2Fw.-s82Y=D8qqđHܶwIEiw|;( Y.OA w؛Θ:#(vVvB}n5zG,V/HgF}a3;rI׿Mg<3Y'{9vQS|LCrw-i%ƕ9#2(JJ_Dm}̏Lۋ{6/Тpѵ1Π Ndr *+@} $%IN ibA^ucykRm儻OèxEpu֪m.Jײң_ zJ˰'=eF[tKx3u]F2^x " X\m^)P8L{L- Nc 2w4i5s_|T^$a.#Ӭ S /-"6{_#'ީR@*r/~|1W?㤞B.<K*5XSMy&t|bwu{2ﰆ[ (yW`7EK%'Ugs皕=3њ!J@l4=T˷@UpX-ϜzJcO3WK6S5cK"-z*R)[$ LH|;@ݲ:뢏](7KNSCk{%~KY(j 5d:>1Wڰwyܲאּ$A/ͬѝ/Qɴ"/[Z#d-R-+dէDadӡ#{JlUm4&H9~\oɇG-!͙nzL psqwu Y[XijlF"]_j}6 {#a5nb7??k~ ˿~ddjޯ_ r?)o%d ԊkKv#L3Kg  Vg #zPv'YQ!J}fosfXIc!dR?&4y{㒜݈LZ¸FUcLags N(j>5ԗp8UIp _} $ELS&kb$$*C{e9.<>V~@oPSĚ`}S ,%-rs:wYO2λ卞_x@0g`ns +\30 YyCu#((*83Zf"y륈4A*Qٰ/6ĩV|b;bZkցAC -P#H羅.05Hܜ 2]_w6ad1 v>Ur4Jьj[{B24ԏvs@J악xBAZp/ 兜tyR.ݥȧk68WL QW4-U:{Td5MyYQ*My܅ōkVti&hzޢ_F;Q.zNBŢi4I@:<[m>?xZ"72Sܱ=SRsG 7-4.W1#tsz ԛ @bV0t $,3JͽMC%+AlYj$ }s݂֜LEcѾ:H-=,)ȋ}3_]6N<:L@gWd֩(HS4Z.19CIJyk G9ZًH,@'#ǃ W"]ed+0SPZc9mVj}$aA* TQD'#㸻4'ji_ݣ_ Wk"~ϯ/wj/waXk(Vm98ѴQdW5fAS==4ʆcoou {炴v|{ؾJ? 4\'e?S"w>5k5?(s>eKA좬4y2&r~t1T"zsWSIq6ҿ؄)^,y(l9Xhۨq}5"H)ri;ٟ1c/=|+7Ec(A8d2ir\>TwXd+Ϙ3o=br$8 B)J6`ӏ n3|aFLX^)vQ=atw~ROn|諘K4N3(;m?ob^вH-+F$Uğo ԁi xipfu};J#V4Jb? cq<`F]&+M@/܅ّg6d XzEoΦSGە,:-m埙7Ar/J[vì̘/JIW50X IBe ML@Oqji{e}yqZVWz[tzO]0L))Q܌~F V44GpIRAs+ dY3B`~uqA+F ⇲ {Wc$zLm2 WAc7R~u 8A>*t->YӐ7K?2CN%e/PiLk gVKLUKz՞G ]Os:sa\8&Pey(YH`%1'8::uA{80ˀe{I^4u['3Pb,K?=~<}y}qlo-|ٔ/c't"G7p^!@EҮ:Jw_~f8%|JE?Dl6?6d|yemi4wz\2kT6-9RF ʖ.t+!iNlvKgB)g?l k΅$IZOBc/rQؿ)Oh pf|DGb=nÿ*%yȞa<)+H&\'%MCr,rޱM s(`RV;Zʴֶ)7Ԗtôǭ1c:ղѱ:/;&5"ڰгgW8<6c}JpA;Ny˦֦1oOo^cy#r85/d0Z?(;^/&rʌ(fg+m\G Jhn跑 {ܧ;!(dSߴ zݠrqi#{`zWZNTV~di^3ҐD)>>4!I?d=K9N$Lc J9&t0HW=mO?Cri+6=EawaAF$!]&H9Nˆ.[rPO "Ph{}#%)VϢHQөnfvmG;vԒD{t{_R]Fȏ0x;5+Ň)}lVR} P.AD< [LXwo֟sN,}"j/kiY*!2]##7:ڕ &øZ35,q8w,gۡϻP(uR4bdzfcCRHRoCSJl9:$@;kuDb={ .NÄy,KqhtgP`q_}!vul@^&Nr8k0AKl)̬` Y8($ = 1? 5 h-gՎI$)[\ҀRI #`q˩AurNGwc\\M(7S BD юT1یt_ !^^$cDž%$]Fn92:;OJ6w'AhAST݊Ϊ[옵J)| şy)ci6mۘi۶mmcӶ5m۶mqg}NrnrdUZU_e'DLW^-e^yB*Ӛӛq@`exH$3i9*dcnQI`eJJV[ ҉PjBҶfQNq.K]˽8\0Tu]x,xԳrI׸edm@H9/y X' rn9( âNgT_zZVRNv{pPAhH&ni݋8 bME4Թ//Rgk%d]5ӆ-PF,Y2g Iau2ƙ4AG !:_q >Ȭ7ڱ3İp⁰ao"٭P󺍊$"DZհ/Գդ$+CjĹ<6+E$j[F6\>7&X2 <܋Ѣ^Q ;kme"Impy*@0 cKP_[KjFKsqxy[ંN oQw[.q!SB3eJp-HSv`|Y97Zc<,q|n`xNDoZ6mp1j%` Ky,cI7(;c@9N-DJ=~R|_Nzr>ܻ岹dz/3'(A)ڇݸ+ގG_3:*u< -< XS*tv$B&BwLq)̺#H=Pc7,lbPƌ&:K.WY2oCl_zG,e!> 'Fv\IdH*?( j̺W}[58W"]>KK9簪"<7m؉05 dN#ldȏf"`G ^S~&O҉\#/g2 0qmԵ 5ńi42cELa =mbk߫?sKX_&jnoBH'o` 0u"dg3đo#!?o ;=-= +!3 -;;4/,7;#%'BKa9QomP702!Mȑ!^^B:߶h ~'/sR ?Nϸ[q X( ߧd&*+LuuL#c{@r)JɃ"}U5V !6A`u(?أ ,rauUr>rԴbv $ǜzy֣2 ܽcj &$ 2Z,>XpIgܶF%OK&DZdۇ$An>+c0#zTn9^C!UHJ2~$V;sVDfN^Xv ٶYID#S4=jyu(*e-dB?G;>< ȭl? ,T°i)xNِ*Of燛}=l=_]0MyEŅs;ϰǫ T@DƖy9]:f:^]g Pr+B[p͌;`kݸ]X skGl+ZcѮ`ԱWUB > 4&¢B)qX)GϽn>ncP nZc@߅_LV ,L+Tsg= f a; !PR[>3-Dv*hF0ibhcY.5VVur OhfZzj}8\?&{MB@?qD$PZq ~Ze he`@`_Ժa`{7- rJcRq4!݅QABcɋ1}/'qnb;w R Ȝ2PJ5.XضZFڎd3:/|S/5u^o<$0$茨:$΍`:)0U,q>HHZ |>NKx>藐LACUECIUiCVSD&ÅBur>(VH~Uh}es.\n]\{fLmF[J&6>oW=׆^Dо[]1P`VDI֤ʈؐGH rNچ_k%peQ%I1I+KXwŏp05u*I (Q$qEWd9$[݅ mKSZ35:%♒F*bϒ=]h)|%? ߚAz{UsDw{y$Dr ~eP kkB+?%+ +pm\5Yc?'82!3:<=>ҰĤU|27ƴܱ57DDDRԅRRyڲȩ)׬L=ֱʏzYzc+:SaBwi #՞G|r bA{qĤf.'?woVqqs1o DBwYIJV֨Ќ ,A'A$u' W6'uI{QG/:tQ8_^.i;=?uyN7g Jt=ߍr)eU/<-tܟ\ҟa,D\%>2%/*!6ϲ?=<1q,pj,Orex⪰(l=~(\I.1; ֍p1b0/fe`_o,!$$hhbLg ;߁,/3IY;Fڌj/? 8Xۙ ӿ?'ǁN/ p6G'*rT k8, 8F'by[-c`hgc`_Q 77``ء\++߇_MllBf'dgb_;9.oߚXX4?,<45#??#_/rWG%u51?$ #hXYYiYbgՁ]%Ff& 8i9Y4V4}]O&fNl FO{f'пcSw ?[ ٙku1[Bff?E׏nL "e? LZv(}g6ɖ񎺍TiS9S@i0A¸1dtW?kb]RHo 4B1?$VU -T,DD(JS_̐?׮S'2︧p4%#=f4ys(uuݍ\ [\VNyW,4@bݽ'S%$-P'Wm6!|`T-R_^!}a؜tהaT}r굙 ~&-Ue"  R=G F|E[ T*7Kxc}^1~W(4vHbv9v6McrOo8d7ɥEXrDwgwqZoUo(@£|qXYETh2Eġ/7#r"a m%vߝfĿ14ӎ`*hksH.W,L$+Io7Vx)cxן4:ĵ(W(arJF;HSC%f.Gk8jBCW()<83.G;L}^c%Ϋ>ރe9t1N gb]@\oC:m-~ K}ni3ֱ?m9ZW##´ˮIQÞoFťf3W4/+1lS "D-s#ó$)+CQ ;hRD3<o,΂m7Ո#+̘xvYk1 ̨ˠgIJsT#eTk\ג."'O=+:sHSߙ+7w.h tqf#c?w((BonD'#J0^4N+=nhtl%%u!f .{6j2yEF:XK-'UP}D#4&' }͵ XsAy%ђ\͵Ѳ³ih, lDeo3ZiYӆn1{Fd-"qNLoj=wi.%PPlLv;uLv'0ZfbAfĝm$ _cy- } yB0'xH2a('c1|xVYEP$-粅RM:&*^; 9pA%be BM|EvO\q':H~F33b^GQ<,9 &Z&ƍe|$dQ3vnlx MX&'\M-WX{;|5΂/grVx̓*%?[2WjeB m9I#sYw/fS&&"u0 0!"TǽpC5`#v"-%$dPGw#n0~sIInjfdż@* G F A,ԟI(%dgbWRҶ3k9 0 P\.8 \EhpxE*+jxjC.f,ufFF "&#_+L(p@0 zE}4)e s^!R[C[6ERi+RFS/yj|Ï,rc;r:ཉ=Uw!P !M ,OCA#w 0W+ufN|^+stVkb`4@O=N=!l\(F -nC@HlڍA"sg& O<~:GȀ:y|`ӯJ!h]'_;%AmZl[qIh; m~M ~(nĐ*1ڷ2gSdž|R'Tw;@Gl.oZ+*#i@AJMx5 jI+/Y)Qs`ZyˉʕcƘm޸b>'d.fooAG5prͻ^1h)W~xJ0/ƙ걨1o~o >ᝲ[F{+e:ϮI"nc#gg_ӱNElk> EE!ْcγSOIlMM_]^^t_mWֆTj1Q\GJmC"O=x:fNNP%VCp|  ^0x({zh_y\뙵K X.fm>Cw% (d\ƚ^H:L:c;=_V_9|Յ{NïuBmF]!95 L"?~(|\! ޱM:ap8k'>BޒU"F .U90[ͻ 'wA6i~=.]A_2:eUYkf̅'piIIaU-L!9q/'>^J<8Ty+朑IJ)g˺.II} [%} 7ܬ 6;Y;RYxtYGP;Ym߾\{}4~sl^G,6̔+ptLQ7KKB&72I-Q2±DK2mlrl8Z8> bؼ׋Uȭ9m[4ـm,iѹLm(nmԽ;ÃEfHw[ҦX畊}<QRhm[6D79Vd@-`n ^W1T{N_Y'c_ƆV<_R ނm,hL? HԲX7=cne RNeL]~/@{򵀀IpP4+1E)hpt?[\=W1tzK! (w55azUǿNIkJ;IJZn)/^e++W̭Z\OH-:`4 Ύ fρ^ASJ,w߿N:qa3=c6dl,r2(BRF@2.~&n*}.DIt{g (ˑtt |^ jjbm5,e&tsp'C٥L- ʏkx|J Oo輍Ĕ,M/u\tp0jǭ=8J/X9?]h!5)9/"ȵ>j}$Js,,$_]Lh0sЙm)Kg`(ަMmÄu}FvN Qw 5G󑰶ji.VcEW8DTrG| mѨ*Yx+&>tjH-x1QqV E̺ ȴm6/-y?r?Y,¯e(0ຳ8T2t[@ yD+!P@jYx6@yPe+"X?L>;عf gpΎP܅FpHV{'$AiV|*hԿj[gfrk6p6ex5nc&Vw}w@/zu Zzp %i9N#!z;;x?zxh eYz<ŵR{>Ѧ hko)tt| 3 jtu([瘶e5B@)7IThnҞvKĠK2Yn#!~(.YKNd܀)lyMԐ1b@<=$>vFhucՈ|4WhH^\mC%1 {8&"m20k!=A<m 8!|vf,Y~Mr8^"cO&i/͈P?D/g~{5wB%rߟ6~QmDfS݋ڋA| ؊ΑXc5Nf-Uy2`)yxvi9Op|kF:jG$tnϕ"'N3LbC/Ct(lhA$zN ˚lwv{Į n\[hI< d<T,,]TX2ێa~'SZWBX_ض&Sе4zmaY(5W+65.hm݈٪sV(.t/_G|xM[}t&c~0 7'|SA.{x@}#zG 'Vݡ=z!~vʱl^?\A`PޝƘcX=^5S-Rzva-GfTz\>FN-w/D4 8 T[va޻l4K#bL؉2w|iN, &T%a`*1/Y/ x^1) 7Ki(a`$qrX#o&2_pv1~1 ,EˇIJ/sc9Uh=NH4} c[*rkJP‘;HӫE2*uS~C_rAh`n 77ƹ5Qc5&ns}-t'b9\n}u4Vv)@l]dq Nt˦~:!ҲbٻS0;_y뙝vi.Zn2_kTdA<4 ]D>ݳ bWd˄ <|;ߡvV ԅ_tH5) Rsk)=nMаrl=sڼذd6 mMާǤ|)jm rV=,Ԇw? Wݼfx{CIܛυ{dA}pK4vѓƺgo~JSmŁ\Wk6Hu8x9h sz ΐxbh_Dm!Tmh ֲ )JQkQ߽&>tT1:695seNN A.Eh9' ӆk#q01qcp%#"w&1} Z5j-Z%cE:G*KděZ"Q6#l:!ֆKyӘT.Xص[nOO^Uu{g}O|"&)+=l"f,"LubI;KWPK}/9e \BhTeIV,DiN$Uw: i9/.֊/g%l ts14cXo21X(u//]l+ Kit 5g;ujQ Q4p |g#y>_;CV`0AL$B9mϷfpF[A5L޺UTՊ!m['!{ 0_6|nAhyTSM=Lm->4QLɿRQL@֪Q[tlPD=21VYa>4_.^gm&]#%T8@jw &vU/k}D '.kݪxQE> uLzU1&U2}Y|gQ⑄ngwMzv޴/mp.x +ůѺ:x,1jFЋG@n@?pOF?PJE[W'D}ONf>+ D$7dctq[Y k9kuvdvZnƜGsƀ~6 pceWޞTS= m~ZyAC>Âl֖2}W fRC1E:r*ֿ~hXLyAkdv YA}KGQ @LDjv0p)֔V0zTh*;%K~6X/H]V@캲F,mB Y<ؒxsto߱,xߒ?y}_cUg2kzAel.CEAM&c¡]\N,e:e#5esY@YW>P 6 A/G@Oߜ4jpwxWXm $fpZlY7/7.OZUnz?:5-e)cB _V/ e_pH18Ԧ^⺼5,m\Btr5*Xtie7=QQ)pMD3ᦃB3 3ױ6suTP%Rb )*NG,1 ` xc t6ևE,~I2oY{: (]7o{v`;q}B={ްAqhċAɄ h1UTk=t` >MCRXUдj״Yy&'D4U͔RNr0e>]T^ш[a9^V -Uw.PVò@ QR~IwOBP*1^hu|1d" !Qy *[WuENF'|Xa--0¡D|X^k-AHN_71/״3_ZD#95M|5_UZ틪OP /"2~Mezfj.l=(V9*99pa8huo͔GTyQ`_~Hs! ߆Gk+"#{ň 4Qo~𡝰Yy;%ߨ2;wgRn.2ꪺI:Gzkz[=:؞su\Au6`ȇl@ˌd@w>T\vSg/r- ΕːUp iH' VYф},u̽WϰD=[fײ 1k}K }tz)6G읜Y$g]V#Ɲs|P))(W gȟy@RL6teZ8[sTA;ov24*[ 1ⴉi bڟU;]β¢6l>(*nm.8u]dp$qte sWu%oz4\0{V[)$~Y^SHղ;t.충y7ǝGlģ$dDPG]&N5OlOȟwcbi(V^  ^ނkN~z3;h/u=ڼh))™xaEAd^緼eCdrqNu#Hޓݽ#*ɂ S~Qn}JdlB[@ \D:`d&q2O"ZU۾lܗ+;DDžk:v5w];SPW]kOPk?J-!zz .\Oq (!sUC]EҦf7O<;'L|ExC" R\Q9Rd ݂5Fđ(p'͐3ޑ(ii~.Co>HSlCrv C=g8́b5j(S=ȃ:UϚh52{'r1[h3$[y_mOӤi kpX͂-VfA/Rx;QjoI yZ$mJYPA')n W> _qK۴\H=O12%'nb!0wzLgݻ94,CGUa\0MPڢ=o,84ǖx#`4p _!&󡦡ӻ3,{<9d 0(SX&{@TFOl.fTBA:^ƬdġI`-(٥kǸΜ*Ed'PLZ#`UTI&|)Ż5.ͷI2u ^z͠󝋥γM3xjy3"7BoMѳY+\Ʀ/XcLJ h**RDm(4IXFƩkl #)-0f1s^\lYuGig=Azɓ8t.N..\.ڝCRuY+,Bwi7%+dތNSx}_0%$o/Bd]s7ۉ +_47&+_Lʗyњ,=`H$O PueڢaSl6ъ+^OM@`~JqԩcxRJaQmDA?UX 5hLKT$gXP *u9V\5ir)RǨ߹!W7B,ܮ.Cz~J̉gw˻S*Vңn]G;{\As]733.e~}pѸu i ?""]浛M,|W.vLWt^qhv?W6UbL}g}ߞunU>qsӈi{ݮvuvޏ[x7nY@A.T[{v[mR'd感Փ^նwnn+9ơʵ(CW9Tss[ ߳Ďhxնzt$vs8Tꭻ*o}|\z{:NDͦSo:wv_q~ | \nOiםˢD4ӣߝRUuPIW;q?}M_ek5hXV>zU>/S?e}^w%n9DoN>I|7/x}8 ~ѳ?&w }mb-e߯-sCܮ2,gɍ{j<\ubP_>}SQBAG {/?V4:t9SmŀMwpRwzϜ5cIGk?5jjtNZX@7ެ WUPUfE>Tyn}T^Ӂg;0SLp`AT#Uk瞝W2*:e }e՚E||ǝnwWpc[Oޯ73;wo^f~ߧ>籹ęG,7հ<>ھש}L|saROh^3b1t}^NZ!j%nYywuu_|tZ.$~']wJ>0v\k;|ia'W?1xʷCߍX+m={{U9Gղ;T~COMtjCr;9s[6_Dy`\wJ/{B~M_\ݼkC їfn'C?T 'E?x᫇)tPA>GNOl|ʒm:i]ۻVs}!=緽+f=iOIEo+8չ)wGxc q.9TkO8r[Q!ݪs|ċ'~xtº.Y}n+Xc lhwlE3[/-骨*>0mʼ &LXo愯 vcwwN34Sqqx]*O q%G'>Z[tڅ&_jΊ-f-ظt{(ۈϊÁJ'<q7)f6K+{٫CC/+t8O W}c#9fd{~s?-~ơy(:T˸7'm 14L/]RVieq G.͜3ԪAY?(mf ڟng8{n.qC;Z%\ sN>m 75]n>6y1W^;TjW/;)6wgzv4oyuogztVW淸87;TpknvV7 -8h^[GfG~ҍ~_UP'm*Z~ӷ6uC'CE)Qh~שʿG&ilJ^pɟ^~\~Q1Gę%5tOφ׉/&yu-/sU{.i{5kZ}6%ڌ)XWMyϚ/gL{Dӟm_i䏺e$=w+b;:jZ^>kJ]i.M"skhg,=ߓ-|Iy=5?X?&L>_r”o'()_S2Or#,Pη=}پ˄/UM]Ђrz==ݩE#Ӟ勩}?sԵ/nmq/? O4/!St=ጕ W.4gR*;~f=m7@:/Z%{GmrU[TvzdU6Fh텇]=w_6n:mvz˅跏jl̼u}閦6{~3> td4I1Dmg|O~qJB.V?;*!d/_Q'}ǭ١*Uv.]}v [-|Rwowq`Woj|SR%؏-)<}V~֎睺"*򨧛kR*i W}g_'Rj齅YW{=0ŝ'x$yeW|g桙k`Lg}C؍=hhN3S&~G*J:<g}?Դ~ZңF2v4'k~צl]k \ճ=Ԍ<.+DO@ūx~8d-Wid+tg-% w:;hu+OӦRA<]}>}+{?aCݼJ'-4Jdd\| zoNݱQOV]24Ǜ.vtqZWirjCC;]w[Pnoh 7w[W'l8풗VSt^mOWnwҲڸ;!ݴ´_=ڝJ4;?#cܲU2ʲM2r$Q=ݻ_Τi/;?{Vm< m/Çm7rX6}_˻-{JI/W}?b%/5˷G7)vnR|8|󛩯{_ZswcKZ0Xoub szU2]?zݟ~=#u'^m_*^T_l^GO1]}^~PϛuE%r2W;R͎3;|&zWrbuzlero/ gֳc~q66\[enާ] n eפ&{4zWj:C6kg5_7bt;/ZSơik«urQnoֹFL>;ׯպg[=T압߳_Ӝ̋ʛDzïG<ɛN{[]h93^߿V6Z+z~jѳn rf[fgϪ2l= f˷/C te#ª. I5>ǰ\wOۇ0#;Ʀ =-s.=,9Ƭܠ{N/_~]ƭZKqbKzZ?՝[(dY.$&y|d6#BF:}|-+V {󴍱ĮCYߦ$]}ris zuɶM6~5β\b\鶱R>MV?c.~V]2q//Ul/ެ^5e6Us=:C*)9ݒz['?ñUFQ _5e* h#Ϸj=:KH؂P?)\SB`]_?o=|}ߋE A,qQ?K;$WcWkQGzն?Z5kVz?_2}ϯv oj׬~j |NMC_lm_g ?WY5 ߫^; 3^?K ?TTAP=uM +@ŴQ<УG}:UUzuׯ_-8ԪW}*" p *gռ} ;F0$ف̏2u5wyWt=ˆaF~ \RmAѦkf_wP"l[mmwgo~oy ~|'xi%DI7 *6NDIDq}>_(I;({AK@cA)4x? iSR`[* cA): O&J%t( 4J`{vQ h,( G>͠ҡPB&&@Ѡ^Ya6!<< nߏ'!^x  Z[(<sXPVKM>#J&Jv">D~.)T4:X@EhhCT6 ~k'zEa~PP$J%8}wh8&`3>XP Ft/t8ӵPP4x? ~X؎J D/q ylkTlOҡ}CH׆{!Qq?T h,lG%=E> = Pp|Gq s6 CsO_:JN!9t8<t8X$JӞ$J% K4 GPӷhhکpLchXP Ft/tҵumQq?T h,lG%}zǿ@;@IX))à@;y:atE4?"~G샊q* Ły4G Pn-^>ﭶ+3<^غw bBTЖ~o%P9- 8F; &| F[xi06g:!;ȳf,ă)ۄc1aL <&?aY1Sw'Mv3h}b||y_o>!&V}pߐf1Gmzp>}`Nq~Nc@Nq^=3)7D-)wfm35~&Ef%b&$p7] ^1x=h#HlutM2X%ȏ)]+6Ń'SgokP@_x2a& =jBA,;.܎ea"LY߇3iY>4;d,)ݘ-݇HĂW]Al z:a.;cfc<=x&)}Ou&O0)PF8F(B&CB^Zۿe(s*I@xm5ÿ-e }Z"!ar$rWʧޗ<> IMpx\/}K+cME3Ae焩QoGk1_},]]#LcyLڴNdܘ ?0fyb3?Y=?&6CTM1O1Yʄ% v`*_ u l]LԵ%ms,l, LfrFyXO)65E\𸼹mij+=c%'VdŊJĆgeEyt1shQUnV "H'-,M) <)<.chh4 y 9SxvQxJ]⪢ yc hZ1x|LJpW!!O!t6<Ձ,:8+n& "nBcj<)󦮾<$ArV܉ajFt_i z_,9 NC:0ÿ.q$LtÃx fX-(LM"vqxǚ#8"qbMÃ8eLtK%xdG.0S+^+p.La=гճxzfk'{DA Ł%E:&ڒ٘!i, &\ K  k&O(<QwP$jBX%8Ңr<'ǭjgWq='KxiuU. &5yykKb~Z[@fDޢ"Qu*'-`WwM"QUQ=/*3t_猴*MԘ3eV .aJ%jaǚfxRx;_I ?~9>%12"%<"ONoo 1Њ2Ȑc g )7cWʒ ߎv J-͓3V28F)kóqaj<)JJOM9n*=)l l[([0gTu1S+<ѷoK M \GKt'Rw.><8mnNv[RXΓ'^H \ Kt=⒂\!GSPkQij1#)#hm۵5z>ҳU-3mƓVO*|yns'ĩ!l]ǚkKb~Z[@ov%lwکsxbXWw v"#v6D9<,+֗7v_NqxXۗ95mAoD^\{At3@Sc<m} ~/m-94lxk3_zVx>7P<|2:-ȕ=l,7I8\;?񣃥CO y sXQ:y@LXw=irssi|\D\F=W{Mxy!6Fy]=Ⳡ9aB:͜%WrɐvRL&΂Ήɐ#^^={e+2#_9祽"c.(O!0$5s;}w!V;rE s& Ht g J }{gWsOZ]hu(/DW#O;wכkgJs!w )}Qar$AY+*}iY.'FĝK}WQGMq fc&3OD^t;1ff1砍9o#6xkʓ60ǡxrߡdӚ9s(sP /s==F~ggOZpPܭrkկH&00f0ӊcn?Z~wIm^Oox{--oKTR{BPR]=XVz^z^F=Yφ&<񡖾>YP宅8P_*J}Z] GOBa52qŰBwU2ͪFPROPZo V [s-iךp=nR2-}mI-w~֭Kcg qzF[J ~ˌx.c=zWz]wU 7iWg-QHAqBȶb# }Giw-,T+/=Z 8CYLGk!fۗJkq}&%zNNqBĎ5͌5uWx{+w]ˎ{1vc*ۮ'wNA9\CtSc3YZw\,7ZV{V+qL6{ Tl8A)3@5 l2(ɳ m֧::9\יg{ܵ-8ew}*)/&po刷ɐ9p^Lbl m#ć5gA9 d2)7ϫ!g4AWAoK5yg e-M.(OB&C0\ cϽp s0:Xn_[u\A+=e: \xg=sxSGkOe7ks wuL]xu;+At0y;s }E#qOĝc+h-tAD{S#"=l_}i}(JMZqeǚ}fxY]uaŽNj1M"J.D91_B"-ꩋqc?ꄙʒ6ǰ+p a.:C,lX3ڜv805IBXs^Bǝcma[|bEH j "\MK#w!ӋpiuUD5@MW=1ľtŗ>ڼ*ՃpɎ!ypδ:ɑ+ț7;40޲,ҳ1  ՕǁnBF8vs4uGMq㮝luʸ =i M'wta=McɱwD5ɠ1Cnh:g'2l]{m֝Ywn h2NX~q8ĶkbE[+Vwnh :MXbu5(] vںeu@2P]Rw.><8ud Ev:8֔{͈] : P/µq8$sk?,0~A:`kĸGޮά+=-}g'3ʌ%y!Z]gkA.Cn wÞl]6SזDfivڥq8r,F[s"Z_޳vMK!q؝Qi})zO!:b9mD]0^ϩic z'K\#qhY9X Dk^EsxcqTĀ_a03H]=u<2*:A~Χc<$w\ńX8GMoZQ/=kJrcl@?OeW4ϫF=gx_Is_ytMS*lWsp(oi$[*L/:v9$|#HzB&CBpx<ʓ8RxG[5 y 9cl W2ay&CM @ mKر8ÊbX<2GXwQf"h^ϩic z"P\!Z c9cm,FkD MDYbSy Z<~5xĘG2\$F1x^5.s2t/k b"!L3q76`~c4%M|"J΍q ۱&FOZ%s}zZ}}lsyh}hP\JT#@#Kv}#C+w)TAdӖ@ţhᨚ!wRM!!O!=7.U7byYj! y 9:Uyj"j  y 9&ǸFNϽ97s0*.߷]X b=XPEl~x`=uwʪzyDiϚk- P"L݉9Ƴl[8W *G"a@1c +F}(D5HF9Fᙪ5SUiD4c5̌52۩ʩ*x{oocc*Y9Dנ >g~m9Vl=ڣVꦙʒ r1UsT{@SU8Xl8XL)Y eg`ۤRPMi `PRqsmy]Ubh4IXTgUb hJ%,T#_پo>Z^j'h J&,U#5\|xڪyhE;POOXڪ9F~pYچ蹷,xd6XøKhl3b <׀9׳jƓVEM5%Q,+qsغh-im+fȓӉ9,Gb<'=(Y1VM1c9sFZ_cRDc[ϩic z'TD396,G;3c<m(9wm̎̌"m,ֳܲ3ܲ9굚ٿٳ@-([+V/%ܵ𴸿դ+=\t7"<ؗ/}5OՓ5ψR = ;5z QKcK 4M\[?ZM 7 w{z_nHSh aRqȁ}EEӞޗzx&w"n~~{Xo?YS(0yqki\K3ֆ4-m-6Ga\>?CƟh,ZמյǩFÖ.ϾѴ+8Aܵdz"ˠ"ѳM98ؗ߳iʹdzջ9+眶ʱnR.7-vBǫ.*ѸҞľԏ,t:ۂ.17ΎzȖ谨ɹw&̸?,~>70&׈ǹ<1d.kvܰ±p~=ǃ%̓%•݉9=e$@+7DZ&~$ܡ#I½%!0̴v<{HxX#Vzf[ {fÌ'k(FWdq7Wu]L][z$WgD,rm xND+{>Qb H%or+֗Tfb~sXf* E~s9ˁbЉ"uW,7cqZBqPpQ2\KԏSVǹ{MԙzfScL cu{xu[0j7. 'ng@i lv:_?Ȋ@Ⱦ' eqҁ|ϐe^(굕1k yJ^X F[՟мBXzgkJ%x}7O諤v(|n} F} 3K EO_EY~nbs>>WYv+bX3? 7So .#%W"BM3cܡvjmQ.nGn1A;;pz͔pr*XYU͔גP?S{ :[y 9pm׊:qr77%`Vǭnp֌uVDC02p& (~2az?Z 0=^h^YFZy&9E#@^UAѹ<{Μ^k s{-a0KoKc/y_Qe=]VThTJ% ;8gxge{/3q1lN4[h2?k"o= =-c4jWq ǒtá*jlsN{菩8ڴ?4+ɲܳү?):+ٴ?GZ#賦4ɴX_ұʎCώC7Aﳦ/,`/"˱6ʲϲ,i06>Գ۲"dz̷yd8ˑ\wzތ9rA-=NcË:@Z";E @;(Daw`sCQ r0݇hIii"n Ӈ_PizeNJY.7 fsC} JD\b9z/&$ ,{rt(Fn'9 ;W"9Tc#sLF4D~l56j\j{98z%*ko?_֎{r.Fn==G>ح#(qѷYkrC}n +׎ 3FB8F1Z!V9^2#;e)S# ssI J(ylj< f0B[Xŗ?C3'- 6n1n1Zg.gYD|b t`3@g^` &ܬy(L`0p&"K1|fS3>1B`80a8'كAQB0/r ȫPsAy[)alFԳb@g(@Mz.666666666vvvvvvvvv66666686666666666666666666666666666666666666666666666666hH6666666666666666666666666666666666666666666666666666666666666666666666666662 0@P`p2( 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p8XVxOJQJ_HmH nH sH tH J`J P{Normal dCJ_H aJmH sH tH Z@Z P{ Heading 1@&m$5CJOJPJQJ\^JaJX@X P{ Heading 2 @&5CJOJPJQJ\^JaJV@V P{ Heading 3d@&5OJPJQJ\^JV@V P{ Heading 4 @&56OJPJQJ\]^JZ@Z P{ Heading 5 @&5B*OJPJQJ\^Jphb@b P{ Heading 6d@&%56B*OJPJQJ\]^JphL@L P{ Heading 7 @&6OJPJQJ]^JN@N P{ Heading 8 @&CJOJPJQJ^JaJX @X P{ Heading 9 @&"6@CJOJPJQJ]^JaJDA`D Default Paragraph FontRiR 0 Table Normal4 l4a (k ( 0No List @@@ P{ List Paragraph ^m$R@R hM0 Balloon Text dCJOJQJ^JaJNN hM0Balloon Text CharCJOJQJ^JaJH"@H A50Captiond5B*CJ\aJphOR1R P{Heading 1 Char5CJOJPJQJ\^JaJ4A4 F*P BibliographyRQR P{Heading 2 Char5CJOJPJQJ\^JaJJaJ P{Heading 3 Char5OJPJQJ\^JPqP P{Heading 4 Char56OJPJQJ\]^JTT P{Heading 5 Char5B*OJPJQJ\^JphZZ P{Heading 6 Char%56B*OJPJQJ\]^JphJJ P{Heading 7 Char6OJPJQJ]^JLL P{Heading 8 CharCJOJPJQJ^JaJVV  P{Heading 9 Char"6@CJOJPJQJ]^JaJV>@V P{TitledPm$@CJ4OJPJQJ^JaJ4HH P{ Title Char@CJ4OJPJQJ^JaJ4VJ@V  HSubtitle Xm$"6@ CJOJPJQJ]^JaJTT H Subtitle Char"6@ CJOJPJQJ]^JaJ*W`* P{`Strong5\PX`!P P{@Emphasis(56@ \]q r:2: P{ No Spacing #d<@< %P{Quote$]h^h6]2Q2 $P{ Quote Char6]`@` 'P{ Intense Quote &P]^a$ 56\]HqH &P{Intense Quote Char 56\]< a< P{0Subtle Emphasis6]>a> P{PIntense Emphasis5\<a< P{Subtle Reference:DaD P{Intense Reference :>*@:a: P{ Book Title 6:@]66 P{p TOC Heading-@& PK!pO[Content_Types].xmlj0Eжr(΢]yl#!MB;.n̨̽\A1&ҫ QWKvUbOX#&1`RT9<l#$>r `С-;c=1g'}ʅ$I1Ê9cY<;*v7'aE\h>=,*8;*4?±ԉoAߤ>82*<")QHxK |]Zz)ӁMSm@\&>!7;ɱʋ3װ1OC5VD Xa?p S4[NS28;Y[꫙,T1|n;+/ʕj\\,E:! t4.T̡ e1 }; [z^pl@ok0e g@GGHPXNT,مde|*YdT\Y䀰+(T7$ow2缂#G֛ʥ?q NK-/M,WgxFV/FQⷶO&ecx\QLW@H!+{[|{!KAi `cm2iU|Y+ ި [[vxrNE3pmR =Y04,!&0+WC܃@oOS2'Sٮ05$ɤ]pm3Ft GɄ-!y"ӉV . `עv,O.%вKasSƭvMz`3{9+e@eՔLy7W_XtlPK! ѐ'theme/theme/_rels/themeManager.xml.relsM 0wooӺ&݈Э5 6?$Q ,.aic21h:qm@RN;d`o7gK(M&$R(.1r'JЊT8V"AȻHu}|$b{P8g/]QAsم(#L[PK-!pO[Content_Types].xmlPK-!֧6 -_rels/.relsPK-!kytheme/theme/themeManager.xmlPK-!!Z!theme/theme/theme1.xmlPK-! ѐ'( theme/theme/_rels/themeManager.xml.relsPK]# G$GJ$8@$,$,  & $ "J L# @(    C c"$? B S  ?K $$$ _Ref165647643 _Ref165706876 _Ref165707370 _Ref165729117 _Ref165729095-Fw A$$vb$$ $ % (  ( ks8$=$?$H$b$$$=N"2wwPV?$$$:::::::: r{3at&q(3i<$65:~Yr?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry F _Data 1TableWDWordDocument48SummaryInformation(DocumentSummaryInformation8CompObj` F Microsoft Word 97-2004 DocumentNB6WWord.Document.8