ࡱ> '` ]bjbj$$ w FFS.$$555P548|u88888u9u9u9|u~u~u~u~u~u~u$xh+{uTCu9u9TCTCu88uVVVTC 88|uVTC|uVVgHj88 X9_5P4j$su0u8j{ T{(jj{\ku9E<VG>?qu9u9u9uuU^u9u9u9uTCTCTCTCj$$j$ Basis for the annotation of EPEC-RolSem Izaskun Aldezabal IXA NLP Group Basque Philology Department University of the Basque Country Donostia, Spain izaskun.aldezabal@ehu.es   Abstract In this paper I present the linguistic resources used when annotating the Reference Corpus for the Processing of Basque EPEC, in terms of semantic roles, argument structure and verb senses (EPEC-RolSem). When facing the annotation at any level, some crucial decisions have to be made, such as the model to be adopted and the criteria for the adaption of such model. Among other reasons, the fact of having the resources I am presenting here has led us to select the PropBank/Verbnet style model (Palmer et al., 2005, Kipper 2005). Concretely, these resources are: the translation of all the verbs in Levin (1993) into Basque and an in-house database with syntactic/semantic subcategorization frames (ssf) for Basque verbs (EADBData Base for Basque Verbs), similar to the mentioned models. By means of the first resource and based on the Levins class, we have linked the Basque verbs with the related PropBank/Verbnet verbs, getting all the corresponding information. On the other hand, the ssf-s of the EADB have been very useful to associate the appropriate English verb sense as well as to define an entry in the PropBank/Verbnet style for the Basque Verb. Introduction In the Ixa research group, the Reference Corpus for the Processing of Basque EPEC is being tagged at many linguistic levels, starting from morphosyntax to semantics, including some pragmatic features (Aduriz et al., 2006). In each level, certain models and tagging manuals have to be developed for the annotators. In the case of semantics, the most difficult task is to establish criteria to define senses in a coherent and understandable way to facilitate the annotation process. Our previous work on semantics, when treating nouns, has mainly focused on Wordnet fine-grained senses (Felbaum, 1998), having as a result the Basque Wordnet (EusWordnet) (Pociello et al, forthcoming) and the Basque Semcor (EuSemcor) (Agirre et al., 2006a). Nevertheless, for the annotation of verbs, this fine-grained orientation has been questioned as some works point out (Ide and Vronis, 1998). This way, our data-base for Basque verbs (EADB) has been built with a more general perspective: although senses are defined for each specific verb, they are thought to be valid across verbs, based on Levins (1993) methodology but mainly following Vzquez et al.s (2000) alternation criteria. Consequently, more coarse-grained senses, similar to cognitive categories, are proposed for each verb entry. In addition, with the aim of defining alternations (either general or language specific ones), the syntactic realizations of the arguments in each sense are also taken into account (see section 2). With all this in mind, the PropBank/Verbnet style model (Palmer et al., 2005, Kipper 2005) was thought to be a suitable one for adopting for Basque, as shown in Agirre et al.(2006b). Other reasons have also persuaded us to choose the PropBank/Verbnet model: The PropBank project starts from a syntactically annotated corpus, as we do. Given the VerbNet lexicon and the annotations in PropBank, many implicit decisions according to problematic issues like argument/adjunct selection for distinguishing each verb senses are settled by examples, and seem therefore easier to replicate when we tag the Basque data. Moreover both (PropBank and Verbnet) resources are complementary as each one has appropriate and different linguistic information for defining verbs and for learning them automatically (Merlo and Van der Plas, 2009). The PropBank model is being developed in other languages such as Chinese (Palmer and Xue, 2003), Spanish and Catalan (Civit et al., 2005a), Dutch (Monachesi et al, 2007), French (Van der Plas et al., 2010) and Russian (Civit et al., 2005b). Having corpora in different languages annotated following the same model makes it possible to carry out crosslingual studies, as it is demonstrated in Korhonen et al. (2010). In the Verb Index, the information regarding PropBank and Verbnet is linked for many verbs. There is also information about other models such as Framenet (Baker et al, 1998), Wordnet (Fellbaum, 1998), Ontonotes (Hovy et al, 2006). This way, we could enrich Basque verbal models with the richer information currently available for English. In this paper I present the main linguistic resources used for the predicate labelling of the EPEC corpus. In section 2, I explain the work carried out to translate the Levins (1993) English verbs into Basque and I show the linking with the PropBank/Verbnet information. In section 3, I describe the syntactic-semantic frames (ssf) used to define verb entries in the EADB and the way of adopting the entry in the PropBank/Verbnet style. Finally, in section 4, the current situation and future work are outlined. Translation of Levins (1993) verbs into Basque and linking them to PropBank/Verbnet Levin (1993) has been a reference to analyze verbs in other languages. She claims that the distinctive behavior of verb classes with respect to the diathesis alternations arises from their meaning: once such a class is identified its members can be examined to isolate the meaning component they have in common. Thus, the diathesis alternations can be used to provide a probe into the elements entering into the lexical representation of word meaning (Levin, 1993:14). Many works have been carried out to compare the alternations she proposes for English with the ones existing in other languages (Jones et al., 1994; Taul, 1995; Saint-Dizier 1995), Basque among them (Aldezabal, 1998, 2004). The studies carried out for Basque show that from the 80 Levins alternations 24 are found in Basque. One reason for that is that some of the alternations in Levin are specific to a few English verbs. It has also to be pointed out that the dialectal variation was not considered, and some works reveal that, the dative alternation (or similar to it) appears in the dialect of the North part (Etxepare and Fernandez, 2011). However, those alternations that occur in specific dialects seem to be more a shade of the meaning of the sentence and they do not seem to be so useful for doing semantic classes (following Levins methodology). Besides, in other languages also occur that many alternations do not exist. However, the alternations that are found in both languages are relevant enough for doing big classes, what we precisely do in our EADB (see section 3). Anyway, when comparing the alternations only the verbs in the examples of the alternations were taken into account. All the verbs are found in the second part of Levins book, where she describes the semantic classes resulting from the shared alternations. Therefore, in order to make a complete comparison, all of them were studied. The first task for that was to identify the equivalent verb in Basque, and then to ensure the differences and similarities both at alternation and class level. Translation criteria and some examples The translations in each class were done based on the Morris dictionary (Morris, 1998) and applying two general criteria: First of all, the meaning of the Levins semantic class was considered. Then, the most syntactically similar equivalent(s) was/were selected. In many cases, the verb in the class and the alternations involved are shared in both languages. For instance, most of the verbs in the 45.1 break verbs class with its prototypical causative/inchoative alternation can be translated without any difficulty; many of the verbs in the 9.1 Put verbs class do not either show any difficulty to be translated. Only we find the fact that for one English verb we can use more than one equivalent. For instance, the verb break in the 45.1 Break Verbs class is translated with the three synonyms hautsi, puskatu, apurtu, because they three mean the same (regarding the class meaning) and admit the causative/inchoative alternation without showing differences at this level. These cases are not problematic since each of the Basque verbs will be linked to the break verb when annotating the corpus. However, as we went going down in the subclasses, some mismatches were found. These are illustrated in this section. - Some of the syntactic properties are shared. For instance, the verbs tell and say which differ between them in the different behavior when admitting the dative alternation (tell admits (I tell sb sth / I tell sth to sb)) while say does not (*I say sb sth), are expressed with the same equivalent verb in Basque: esan. In Basque, there is not a different verb that reflects this syntactic variation, and the valence properties are the same as in the both English verbs. In those cases, both English verbs should be assigned to the Basque esan verb when annotating the corpus. - A single word is used in English while two are necessary in Basque. These are mostly those verbs that have a manner or an instrument meaning incorporated, such us many of verbs into the funnel verbs, wipe verbs, spray/load verbs, drive verbs, poison verbs, and verbs of instrument of communication, among others. E.g.: ladle: burduntzaliaz bota, literally meaning to throw with a ladle. In these cases it may happen that the concept expressed by the verb that lexicalizes the manner in English is not a lexicalized concept in Basque. The example above represents that case. Therefore, these verbs should have to be considered and analyzed into the single verb (bota throw) class (17.1), where manner is going to be a possible adjunct. These cases are more difficult to solve when annotating: the annotator should realize that the non lexicalized adjunct + verb expression in Basque has to be annotated with the appropriate single word in English. - The same verb is used in Basque for different verbs in different classes, but the object candidates must be specified as in English. For instance, Verbs in the 13.4.2 Equip verbs class in Basque are translated with the same verb as in 13.1 give or 13.2 contribute verb classes (for example, charge (13.4.2): ardura eman -> lit. to give (13.1) the charge). That is, in Basque such distinction does not exist (to charge somebody with a task but not *to give somebody with a task). However, in order to provide the equivalent for charge, the object (zeregin baten ardura: charge of a task) must be equally specified in Basque. In these cases, when annotating the corpus the verb charge should be used when in the Basque sentence appear give the charge (of a task) and it should also be considered a multiword. Linking to PropBank/Verbnet Taking into account all these phenomena, we are able to say a priory that when linking the Propbank/Verbnet equivalent to the Basque verb, the argument structures (at least at valence level) of English and Basque verbs are not going to be the same in some cases, and, as a consequence, neither the alternations involved on them. Moreover, when the concepts are not lexicalized in Basque, there will be an element that will be appearing as an apart adjunct (and not as an argument) in Basque, while in English there will not be a syntactic counterpart (but it will be incorporated in the verb). In any case, the information obtained from the linking regarding the sense and rolesets will be very helpful in the process of building the Basque verb entry with the PropBank/Verbnet scheme (although classes are not shared). In table 1 a list of some of the verbs after the linking based on Levins classes is shown. glue22.4erantsi, kolatugo47.7joango51.1joangobble38glu-glu egingobble39.3irentsiTable 1: the link between the PropBank and Basque verbs based on Levins (1993) class. The EADB: data-base of syntactic-semantic frames (ssf) for the Basque verbs Following the methodology that Levin suggests in her work, the crucial task is to detect those alternations that are semantically sensitive and then find the semantic components that would be in the lexical representation of the verbs. For this task, and taking a revised point of view of the alternation concept which is also adopted in other works (Vzquez et al., 2000; Rebolledo, 2002), I studied 100 Basque verbs basing on real corpus examples (Aldezabal, 2004). I concluded that each verb has one or more prototypical frames to express any of the general semantic values appearing when analyzing verbs in general. These semantic values are not senses in the way they appear in the dictionaries, but basic cognitive categories or general predicate types which can serve to propose big classes at semantic level (such as change of state, change of position, activity of an entity, creation of an entity, assignment of an attribute, exchange of an entity, situation of an entity and so on). This semantic information is expressed by general semantic roles (or semantic components) coherently combined (that is, for a verb to express the general predicate change of state, at least an affected theme must be contain; or for a verb to express the general predicate creation of an entity, a created theme must be contain, and so on). This way, some verbs share the capability to represent the same general predicate. However, it does not mean neither they should have the same syntactic frames (although it happens in many cases), nor they share the same alternations (although it also happens in many cases). Based on that assumption, I defined a number of syntactic-semantic frames (ssf) for each verb. Each ssf is formed by semantic roles and the declension case that syntactically realizes this role. The ssf-s that have the same semantic roles define a verbal coarse-grained sense and are considered syntactic variants of an alternation. Different sets of semantic roles reflect different senses. This is similar to the PropBank model, where each of the syntactic variants (similar to a frame) pertains to a verbal sense (similar to a roleset). In Table 2 we can see an example of the ssf-s for the verb esan. It has two senses and the first one contains two syntactic variants. The first sense can be translated as tell/say as in Levins 37 Verbs of communication class, and the second sense as call, as in Levins 29 Verbs with Predicative Complements class. esan-1 (= tell/say): Activity (communication) of an entity. Two arguments in two syntactic variants: esan-1.1: arg1_ERG, arg2_ABS esan-1.2: arg1_ERG, arg2_COMP esan-2 (= call): Assignment of an attribute. Three arguments in a single syntactic realization: esan-2: arg1_ERG, arg2_ABS, arg3_DAT Table 2. Syntactic-semantic frames for the verb esan (=tell/say/call) as provided by the EADB lexicon. These ssf-s together with the information obtained from the link to PropBank/Verbnet are a robust basis to define the new lexical entry with the PropBank/Verbnet scheme and to go on tagging the EPEC corpus in such framework. Table 3 shows the adopted PropBank/Verbnet entry for the verb esan. Basque verb: esansay.01/tell.01call.01Arg0: Agent (ERG) Arg1: Topic (ABS/COMP) Arg2: Recipient (DAT) Arg3: Attributive (INS/-i buruz/ -i erreferentzia eginez/...)Arg0: Agent (ERG) Arg1: Theme (DAT) Arg2: Predicate (ABS) Table 3: The PropBank/Verbnet style entry for the verb esan. Current situation and future lines We have already annotated a sample of sentences for each of the 100 verbs including in the EADB. During the tagging process some adjustments had to be made, because of differences both at multiword level and at valence level. For instance, in some verbs of motion an extend argument is defined for the English verb while in Basque it does not exist. Besides, the annotation has been evaluated and one of the most significant conclusions has been that before annotating, taggers must clearly understand the entries that have been adapted to the PropBank/Verbnet model. In addition, it must be also taken into account that multiword expressions are problematic and that it is necessary to decide what to do with those cases. Moreover, in order to avoid confusions with modifiers, it is important to provide some information or guidelines, although we know that some things will remain unsolved since they are subjective. At present, we are planning to automatize the annotation-process taking into account the lexicon resulting from the annotated corpus. As a first step, we will detect the univocal role_case pairs, and then we will automatically annotate the occurrences of the corpus, including its corresponding verb sense. For the automatic annotation of new verbs, class based cross studies will be carried out. References Aduriz I., Aranzabe M.J., Arriola J. M., Atutxa A., Daz de Ilarraza A., Ezeiza N., Gojenola K., Oronoz M., Soroa A., Urizar R. 2006. Methodology and steps towards the construction of EPEC, a corpus of written Basque tagged at morphological and syntactic levels for the automatic processing. Corpus Linguistics Around the World. Book series: Language and Computers. Vol 56 (pag 1- 15). Andrew Wilson, Paul Rayson, and Dawn Archer. Rodopi (eds.). Netherlands. Agirre E., Aldezabal I., Etxeberria J., Iruskieta M., Izagirre E., Mendizabal K., Pociello E. 2006. (Agirre et al 2006a). A methodology for the joint development of the Basque WordNet and Semcor. Proceedings of the 5th International Conference on Language Resources and Evaluations (LREC). Genoa (Italy). Agirre E., Aldezabal I., Etxeberria J., Pociello E. 2006. (Agirre et al 2006b). A Preliminary Study for Building the Basque PropBank. Proceedings of the 5th International Conference on Language Resources and Evaluations (LREC). Genoa (Italy). Aldezabal I., Aranzabe M., Daz de Ilarraza A., Estarrona A. 2010. Building the Basque PropBank. Nicoletta Calzolari, Khalid Choukri, Bente Maegaard, Joseph Mariani, Jan Odjik, Stelios Piperidis, Mike Rosner and Daniel Tapias (eds.). Proceedings of the Seventh Conference on International Language Resources and Evaluation (LREC 2010). European Language Resources Association (ELRA). Valletta. Aldezabal I. 2004. Aditz-azpikategorizazioaren azterketa. 100 aditzen azterketa zehatza, Levin (1993) oinarri harturik eta metodo automatikoak baliatuz. PhD thesis. Basque Philology Department Leioa. UPV-EHU. Aldezabal I. 1998. Levins verb classes and Basque. A comparative approach. Colloquium series of the Department of Computer Science at UMIACS (University of Maryland Institute for Advanced Computer Studies). Baker, C.F., Fillmore, C.J., Lowe, J.B. (1998). The Berkeley FrameNet project. In Proceedings of the COLING-ACL. Montreal, Canada. Etxepare, R. and Fernandez B. (to appear in 2011). Microparameters in Dative Constructions. Oxford Studies in Comparative Syntax. Oxford University Press. Under contract. Fellbaum C. 1998. Wordnet, an electronic lexical database. MIT Press. Hovy E., Marcus M., Palmer M., Ramshaw L. and Weischedel R. 2006. OntoNotes: The 90% Solution. Proceedings of HLT/NAACL. New York. Ide N. and Vronis J. 1998. "Word Sense Disambiguation: State of the art". Computational Linguistics, 1998, 24 (1). Jones D., Berwick R., Cho F., Khan Z., Kohl K., Nomura N, Radhakrishnan A., Sauerland U. & Ulicny B. 1994. Verb Classes and Alternations in Bangla, German, English, and Korean. Massachussetts Institute of Technology center for Biological and Computational Learning and the Artificial Intelligence Laboratory. Kipper, K. 2005. Verbnet: A broad-coverage, comprehensive verb lexicon. Ph.D. thesis, University of Pennsylvnia. Levin B. 1993. English Verb Classes and Alternations. A preliminary Investigation. Chicago and London. The University of Chicago Press. Korhonen A., Sun L, Poibeau T. and Messiant C. 2010. Investigating the cross-linguistic potential of Verbnet-style classification. Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010). Beijing. Merlo P. and Van der Plas L. 2009. Abstraction and Generalization in Semantic Role Labels: PropBank, Verbnet or both? Proceedings of the 47th Annual Meeting of the ACL and the 4th IJCNLP of the AFNLP. Suntec, Singapore. Monachesi G., Stevens G., Trapman J. 2007. adding semantic role annotation to a corpus of written Dutch. Proceedings of the Lingusitcs Annotation Workshop (LAW). Prague, Czech repuplic. Morris M.. 1998. Morris Hiztegia. Palmer M., Gildea D., Kingsbury P. 2005. The Proposition Bank: A Corpus Annotated with Semantic Roles. In Computational Linguistics Journal. 31:1. Pociello E., Agirre E. and Aldezabal I. (to appear in Language Resources and Evaluation (LRE)). The Basque Lexical Knowledge Base: the Basque Wordnet. Rebolledo M. 2002. Estructura sintctica y significado verbal. Dissertation. University of Santiago de Compostela. Saint-Dizier P. 1995. A semantic classification of French verbs based on B. Levins approach. Research report. IRIT. Taul M. 1995. Representacin verbal en una Base de Conocimiento Lxico. PhD. Barcelona Van der Plas L., Samardzic T. and Merlo P. Cross-lingual Validity of PropBank in the Manual Annotation of French. Proceedings of the Fouth Lingusitic Annotation Workshop, ACL 2010. Uppsala, Sweden. Vzquez G., Fernndez A., Mart M. A. 2000. Clasificacin verbal. Alternancias de ditesis. Quaderns de Sintagma 3. Edicions de la Universitat de Lleida.  ixa.si.ehu.es  http://verbs.colorado.edu/verb-index/index.php  It has to be noted that the Levins classes have been revised in PropBank/Verbnet. Consequently some verbs remained without any link (Aldezabal et al, 2010).  I propose 13 general predicates and 21 semantic roles in total.  ERG: ergative declension case; ABS: absolutive declension case; COMP: completive clause; DAT: dative declension case.  The instrumental declension case  These are complex prepositions meaning regarding, with respect to     "$%'*;<=Kɽrg\UNDhfCJnHtH h9- hG h,hGh,h2 nHtHh(.h2 nHtH$h(.h(.CJNHOJQJ^JaJ h(.h(.CJOJQJ^JaJh(.CJaJh(.h(.CJaJh4h(.CJaJh45CJaJh(.h(.5CJaJ'h(.hXS5CJaJmH nHsH tHh9- hGnHtHh*`nHtHh; nHtHhqnHtH()*=Kg $IfgdGODkd$$IfTl t644 laT $Ifgd]} 2d$Ifgd@Sx 2d$Ifgd(.gdG gdGgdG[]]L Y 'tr/ & F *$^`gd(aFgd(aF # & Fgd}S4#gdG]gd]} & FgdG gdGKkdM$$IfTFO00 u4 E4 FaBT ? @ [ o p     ) * 0 1 C D G ̿tt́ttjj̋``h-ECJnHtHh!CJnHtHh`RhCJnHtHh?sCJnHtHh`RhzCJnHtHh`RhuCJnHtHh`Rh*`CJnHtHh`RhlCJnHtHh`RhICJnHtHh`RhZ'CJnHtHh`RhF CJnHtHh`Rh CJnHtHh`RhqCJnHtH%G I J ] g { | ~    ! " & ̿󞑞󿃿vi\R\E\h`Rh>2CJnHtHh?sCJnHtHh`Rh{GCJnHtHh LhCJnHtHh Lh`RCJnHtHh`Rh`R6CJnHtHh`RhCJnHtHh`Rh CJnHtHh!CJnHtHh`RCJnHtHh`Rh`RCJnHtHh`RhCJnHtHh`RhICJnHtHh`RhzCJnHtHh`RhZ'CJnHtH& ' 3 4 8    $ & 2 ¸}pc}V}V}VLV}V}hbCJmH sH hWyhKDCJmH sH h Uh!CJmH sH h UhKDCJmH sH hWyh!CJmH sH h!aJmH sH h1aJmH sH h1h1CJmH sH h1h`RCJnHtHh`RCJnHtHh`RhZECJnHtHh`Rh{GCJnHtHh-ECJnHtHh`Rh CJnHtHh`Rh>2CJnHtH2 6 7 ^ ` h l m ~   6 7 K L Y d m r ٵ󵨵Ϟ”χsfsh;Oi5CJaJnHtHh 5CJaJnHtH h9- hGhWyhGCJnHtHhWyCJnHtHh:hCJmH sH h13h13CJmH sH h13hKDCJmH sH hWyhWyCJmH sH hWyCJmH sH hWyhKDCJmH sH h13h(o-CJmH sH h13h!CJmH sH !r s t u  "*079HJKPµµ›q_qқPChN5CJaJnHtHh 5CJNHaJnHtH"hqL~hqL~56CJaJnHtHhqL~hqL~5CJaJnHtHh/D5CJaJnHtHhF 5CJaJnHtHh 5CJaJnHtHhqL~5CJaJnHtHhWy5CJaJnHtHh>h 5CJaJnHtHh 5CJaJnHtHh;Oi5CJaJnHtH&jh 0J5CJUaJnHtH56;Y[_ɼ墒s_K:!h135CJaJmH nHsH tH'h;Oih}S45CJaJmH nHsH tH'h;OihuH5CJaJmH nHsH tHh!#5CJaJnHtH#h;Oih;Oi5CJNHaJnHtHh;Oih;Oi5CJaJnHtHhuH5CJaJnHtHh135CJaJnHtHh 5CJaJnHtHh;Oi5CJaJnHtHh 5CJNHaJnHtHh 5CJaJnHtHhN5CJaJnHtH_`deho뭙q[qJ6'hz6hz65CJaJmH nHsH tH!h9h5CJaJmH nHsH tH+h;Oih U5CJNHaJmH nHsH tH'h;Oih U5CJaJmH nHsH tH'h>@Kh>@K5CJaJmH nHsH tH'h>@Kh}S45CJaJmH nHsH tH'h;OihuH5CJaJmH nHsH tH+h;Oih @5CJNHaJmH nHsH tH'h;Oih @5CJaJmH nHsH tH'h;Oih}S45CJaJmH nHsH tH $;<=Ka|}ƲƲƞwaP@h;Oih}S45CJaJnHtH!hxA5CJaJmH nHsH tH+h;Oih}S45CJNHaJmH nHsH tH'h;OihJ)5CJaJmH nHsH tH$h;Oih}S4CJaJmH nHsH tH'h;Oih}S45CJaJmH nHsH tH'h;Oih U5CJaJmH nHsH tH'h;OihuH5CJaJmH nHsH tH!h9h5CJaJmH nHsH tH'hz6h9h5CJaJmH nHsH tH+,-4?FOPX]{|ʺxʊʊhXhhHhh13h135CJaJnHtHh13hv5CJaJnHtHh13h|j5CJaJnHtH#h13h U5CJNHaJnHtHh13h U5CJaJnHtHh13hxA5CJaJnHtHh13huH5CJaJnHtHh13hc>5CJaJnHtHh13h}S45CJaJnHtH'h13hv5CJaJmH nHsH tH!hv5CJaJmH nHsH tHݽݍ}m}mm}]M=h;Oih @5CJaJnHtHh;Oih 5CJaJnHtHh13h 5CJaJnHtHh13huH5CJaJnHtHh13h @5CJaJnHtHh13h}S45CJaJnHtHh13hb5CJaJnHtHh13hv5CJaJnHtHhVh|j5CJaJnHtHh13hb5CJaJnHtHh13h|j5CJaJnHtH#h13h|j5CJNHaJnHtH:;H\]^k  "$BC\ЖУ||Уm`|Q|h @5CJNHaJnHtHh5CJaJnHtHhuH5CJNHaJnHtHh @5CJaJnHtHh U5CJaJnHtHhJ)5CJaJnHtHhuH5CJaJnHtHh;OihuH5CJaJnHtHh;Oih(aF5CJaJnHtHhv5CJaJnHtHh;Oih @5CJaJnHtH#h;Oih @5CJNHaJnHtH\dj~'ööЦÆvfv\R\hvaJnHtHh(aFaJnHtHhVhV5CJaJnHtHhVh(aF5CJaJnHtHhVh @5CJaJnHtHh13hv5CJaJnHtHh13h(aF5CJaJnHtHhv5CJaJnHtHh @5CJaJnHtHh(aF5CJaJnHtH"h @h @56CJaJnHtHh @h @5CJaJnHtH't+:AB]gڴ{ڴhڴUB$h~{[h9-dCJaJmH nHsH tH$h~{[hCJaJmH nHsH tH$hrnChrnCCJaJmH nHsH tH$hrnChJ)CJaJmH nHsH tH$hrnCh >CJaJmH nHsH tH$hrnChP7CJaJmH nHsH tH$hrnCh ,CJaJmH nHsH tH$hrnChvCJaJmH nHsH tH$hrnCh(aFCJaJmH nHsH tH$h~{[h(aFCJaJmH nHsH tHAH^`aejǴǡpdUFUd7h~{[hCJaJmH sH h9hh13CJaJmH sH h9hh ,CJaJmH sH hICJaJmH sH h~{[h5SCJaJmH sH h~{[h ,CJaJmH sH $h~{[h ,CJaJmH nHsH tH$h Lh(aFCJaJmH nHsH tH$h LhcCJaJmH nHsH tH$h Lh9-dCJaJmH nHsH tH$h~{[h9-dCJaJmH nHsH tH$h~{[h5SCJaJmH nHsH tHtaUW!##R$$$-((*gd/4gd^ & Fgd/4`gd%gd,# & F`gd=# & F`gd%I # & Fgd%#gd0 # & Fgd}S4/ & F d*$^`gd~{[ .3PQR%ʷʤ~n[L:.LhoCJaJmH sH #jh 0JCJUaJmH sH h~{[h6QCJaJmH sH $h~{[h6QCJaJmH nHsH tHhJ!CJaJmH nHsH tH$h~{[h6QCJaJmH nHsH tH$hohoCJaJmH nHsH tH$h~{[h(aFCJaJmH nHsH tH%h~{[h ,B* CJaJmH ph3sH h~{[h ,CJaJmH sH h9hhoLCJaJmH sH hoLCJaJmH sH hJ!CJaJmH sH %NQ\fw~Ǹөpp]N?Nhk?h*okCJaJmH sH hk?hoLCJaJmH sH %h~{[h6QB*CJaJmH phsH h*okhK CJaJmH sH h*okh6QCJaJmH sH hoLh6QCJaJmH sH h[CJaJmH sH hoLhoLCJaJmH sH h~{[hoCJaJmH sH hoCJaJmH sH h~{[h6QCJaJmH sH h13hoLCJaJmH sH h13h6QCJaJmH sH TUcdelϼr`rPC6hD5CJaJnHtHh/D5CJaJnHtHha{|h 5CJaJnHtH#ha{|hoL5CJNHaJnHtHha{|hoL5CJaJnHtHh 5CJaJnHtHh e5CJaJnHtHh 5CJaJnHtH$h~{[h(aFCJaJmH nHsH tH$h~{[h6QCJaJmH nHsH tH%h~{[h6QB*CJaJmH phsH hoLh6QCJaJmH sH hk?h6QCJaJmH sH  &+-67:=>cfpqrs~򱤾嗊}pc}c}c}hI5CJaJnHtHh e5CJaJnHtHhv5CJaJnHtHh3n5CJaJnHtHh/D5CJaJnHtHha{|5CJaJnHtHhoL5CJaJnHtHhR5CJaJnHtHhD5CJaJnHtHh 5CJaJnHtHh 5CJaJnHtHh:+5CJaJnHtH#  &-HKLTUVWfu{ؾخءؑ}y}yun^RFRFRh5h>25CJaJh5hF 5CJaJh5hF 5CJaJnHtH h9- hGhQ:hDh h5h}S45CJaJnHtHh}S4h}S45CJaJnHtHh e5CJaJnHtHhK|ShK|S5CJaJnHtHh/D5CJaJnHtHhD5CJaJnHtHh 5CJaJnHtHhv5CJaJnHtHhP5CJaJnHtH23iprsoyoboREh%5CJaJnHtHh5h>25CJaJnHtHhD5CJaJnHtHh5h 5CJaJnHtHha{|ha{|5CJaJnHtHha{|h5CJaJnHtH#h5h5CJNHaJnHtHh5h5CJaJnHtHh5hp5CJaJnHtHhHu5CJaJnHtHh5hF 5CJaJh%5CJaJh5h%5CJaJ789?@EFJLUV[个业~pbTbha{|h50J5\aJha{|h\!0J5\aJha{|hD0J5\aJha{|h%0J5\aJhfhv/0J5\aJhfhQ0J5NH\aJhfh/D0J5\aJhfhQ0J5\aJhfhF 0J5NH\aJhfhD0J5\aJhfhF 0J5\aJhfh%0J5\aJ[abcdklp-JKO^ghoz̾yk`k`kUkhv]0J5\aJhHu0J5\aJhfhf0J5\aJha{|ha{|0J5\aJhf0J5\aJhfh3 0J5\aJha{|hHu0J5\aJha{|h3 0J5NH\aJha{|h3 0J5\aJha{|hv/0J\aJha{|h%0J\aJha{|hv/0J5\aJha{|h@*{0J5\aJ   ' F W Y e f } ~ xjxj\N@N@Nhv]hv]0J5\aJhv]h!0J5\aJhv]hP>0J5\aJhv]h5H0J5\aJhv]h3 0J5\aJha{|h3 0J5\aJha{|h!C0J5\aJh9hhv]0J5NH\aJh9hh9h0J5\aJh9hhv]0J5\aJhfh!C0J5\aJh<20J5\aJh!C0J5\aJha{|hgv0J5\aJ~ !!$!*!+!4!>!ƶtttdTDhGWahv5CJaJnHtHhGWah5CJaJnHtHhGWahf5CJaJnHtHhGWah15CJaJnHtH#hGWah3 5CJNHaJnHtHhGWah3 5CJaJnHtHhGWaha{|5CJaJnHtHhGWah%I5CJaJnHtHhGWah!5CJaJnHtHhv]h5H0J5\aJhv]h%I0J5\aJhv]ha{|0J5\aJ>!?!D!j!m!n!r!s!!!!!!!!!!!!!!"7"ͽ͛͛{k[KkKkKhGWah3d5CJaJnHtHhGWahY5CJaJnHtHhGWahr5CJaJnHtHhGWah5CJ\aJ#hGWahp5CJNHaJnHtHhGWahp5CJaJnHtH#hGWahp5CJNHaJnHtHhGWahp5CJaJnHtHhGWah5CJaJnHtHhGWahv5CJaJnHtH#hGWahv5CJNHaJnHtH7"8":">"""""""""#### #(#)#t#u######## $Ͻ߭ߑwkd`YPJP h0Jh/Dh90J hhhaD h^Nhb/hGWah5CJaJhGWah0J5NHaJhGWah]B0J5aJhGWah0J5aJhGWah]B5CJaJnHtHhGWahz5CJaJnHtH#hGWah7$+5CJNHaJnHtHhGWah7$+5CJaJnHtHhGWah5CJaJnHtHhGWahE5CJaJnHtH $$$$$<$J$K$R$x$z$$$$$$$$$$$$%%%>%?%M%O%h%m%n%t%y%z%{%%%оиzpg` hY0J6hYhb0Jhbhb0J6 hY0J hb0Jh1h"^ 0Jh1hN0J hN0J h"^ 0Jh0JNH h}a0J hG0J h0J hF0J hD0J ht0J h90Jh/Dh90JNHh/Dh90J h"0Jh"h.0J6 h.0J%%%%%%%%%%%%%%%%%%%%%%&& & &&*&+&C&&&&&&&&ȾȵѦ㯠~xodo[oh}ahv]0Jh}ahv]0JNHh}ahv]0J hv]0Jh1h-0JNH hY0J6hYh2Y0J h2Y0J6 hY0Jh1hft0J h2Y0Jh1h-0JhD0JB*phh1hD0Jh1h}a0Jh1h,0Jh1hN0JhN0JNH h,0J hN0J#&&&&&'' '''9':'g'h''''''''( (*(+(,(-(<(@(A(K(L(](^(_(b(d(n(ɾ쵫晐xoi`h1h]0J h,0Jh1h,0Jh1h10Jh1hoX0J h2Y0Jh1hJe0Jh1hnT0Jh1hv]0Jh;hv]0J6h;hv]0JhnYhv]0JNHhnYhv]0Jh}ahv]0JNHh}ahv]0J hv]0Jh}ahv]0Jh}ahv]0J6%n(o(((((((((()))*)+)/)I)R)V)Y)~))))))))))))))ǺԫԢԂ|Ԃtn_X h90J6ht h90JOJQJ^J hmL0Jh90JNH h}\0J h/40J hyO0Jh/4h90Jh/4h/40JNHh/4h/40Jh=;h90JOJQJ^Jhuh90JOJQJht h90JOJQJ h90J h_0J h}a0Jh^h,0J h,0Jh]0JB*ph"))))).*4*7*C*^*a*m*n*r*s*x*y*z****************++++)+++˼zttltth_0JNH h_0J hoX0JhYh/40JhYhcH90JhYhp0JhYh;0JOJQJ^JhYh;0JhYhoX0JhYh}a0J hmL0J h}a0Jh/4h/40Jh/40JNH h/40J h?0J ht 0J hyO0J h90J&++>+A+B+G+o+p++++++++++++++++++++++++++++++,,,,,,*,1,2,7,9,G,K,L,M,Яht ht 0JmH sH h}\0JOJQJ^JmH sH ht h{eI0JmH sH $ht h{eI0JOJQJ^JmH sH haK0JNH haK0J ht 0Jh/4h{eI0J6 hu0JhM2h{eI0JNH hyO0JhM2h{eI0J2M,P,^,_,c,h,q,r,v,w,x,y,},,,,,,,,,,,,, - --&-'-(-,-1-2-3-<->-I-J-|vpjah(rh||0J h||0J hG0J h(r0Jh;h/G0Jh;hyO0Jh;hr0Jh;h00Jh-lhaK0JNH hu0J hyO0Jh-lhaK0J h00J ht 0JhM2h{eI0JNH h}\0JhM2h{eI0Jht ht 0JmH sH ht h90JmH sH &*x,.11G3P455555 /$If`gd gd3 # & Fgdu %B^B`gdgd gd-lgdt J-U-t-}------------......///-/4/5/0Jhs0JNH hs0J hP0J hvo0J hY0JhYhaK0JhYhY0JhYhoX0JhYh||0JOJQJ^JhoX0JOJQJ^Jh}\h}\0JOJQJ^J hG0Jh-lhaK0J hoX0J R/W/^/d/w/x/}/~//////////////////0Ƽ䲼{naTJh90JmH sH h9h 0JmH sH hvoh 0JmH sH hvohP0JmH sH hP0JOJQJ^JmH sH $huh 0JOJQJ^JmH sH h 0JmH sH hJ>0JmH sH hs0JmH sH h| 0JmH sH  hs0Jh9h90JNHmH sH h9h90Jh9h90JmH sH h4h90J6mH sH 00 000)0:0H0I0t0|00000000000000ŸŮ|obXNhc?0JmH sH h@0JmH sH h(rh@0JmH sH h(rhc?0JmH sH h(r0JmH sH hJ>0JmH sH h5)g0JmH sH h| 0JmH sH hs0JmH sH h}0JmH sH h(rhP0JmH sH h(rh}0JmH sH h(rhs0JmH sH h(rh 0JmH sH h 0JmH sH hP0JmH sH 0000000000001111111'1.1/101G1ɵɏrdZPZF0JmH sH h5)gh2(0J>*mH sH h2(h2(0JmH sH h2(0JOJQJ^JmH sH $huh2(0JOJQJ^JmH sH $h2(h2(0JOJQJ^JmH sH 'h5)gh2(0J>*OJQJ^JmH sH h2(0JmH sH hc?0JmH sH hJ>0JOJQJ^JmH sH $hJ>hc?0JOJQJ^JmH sH G1L1R1[1\1b1q1}111111111111111Ϸtg]PLDhb/mH sH hh h90JmH sH hvo0JmH sH h7Mh*0u0JmH sH h7Mh7M0JmH sH h7Mh5)g0J6mH sH h*0uh5)g0J6mH sH h*0uhP0J6mH sH hP0JNHmH sH hP0JmH sH h7MhP0J6mH sH h7MhP0JmH sH h7Mh00JmH sH h7Mh5)g0JmH sH h5)g0JmH sH 12222292A2B2G2g222222222222333Ĵ㧚֊zjXK>KKh5CJaJmH sH h05CJaJmH sH #h0hXq5CJNHaJmH sH h0hFs5CJaJmH sH h0h}_5CJaJmH sH h0hXq5CJaJmH sH h.JH5CJaJmH sH hG5CJaJmH sH h0hu5CJaJmH sH "h0hu56CJaJmH sH h}_5CJaJmH sH h0hw85CJaJmH sH h0hw80J5aJ33 3#35363:3<3D3E3F3G3Q3Y3Z3j3k333333333334幩xxkkk^kQQkh0h}_aJmH sH h0hFsaJmH sH h0h3aJmH sH h0h-laJmH sH h}_aJmH sH h7MaJmH sH h0h35CJaJmH sH h0haK5CJaJmH sH h0hXq5CJaJmH sH h}\5CJaJmH sH h05CJNHaJmH sH h05CJaJmH sH h}_5CJaJmH sH 4 4*454O4P4[4]4m4s4440555:5=5k5555555555555555556ȾȴȆ{o{o{o{o{c{ZhN`hN`CJhN`hN`6CJaJhkbhN`6CJaJhkbhN`CJaJhb/aJmH sH jhN`0JUaJmH sH h>@KaJmH sH hN`aJmH sH h9haJmH sH ha4=aJmH sH hGaJmH sH h0haJmH sH h}_aJmH sH h0h3aJmH sH h}\aJmH sH "55555k^^^ /$If`gd kd$$IfTlF@  t0o 6    44 laT55555k^^^ /$If`gd kd=$$IfTlF@  t0o 6    44 laT55555k^^^ /$If`gd kd$$IfTlF@  t0o 6    44 laT55555k^^^ /$If`gd kd$$IfTlF@  t0o 6    44 laT55L6M667m8<k^YTHC>gdgdjR # & Fgd##gdmvgd3 ^`gdpOkd$$IfTlF@  t0o 6    44 laT6'6,656;6=6E6J6K6L6M6e666666666666666677Žŵ͵zjZjH#hjh#5CJNHaJnHtHhjhj5CJaJnHtHhjh#5CJaJnHtHhjhj0J5aJhjh#0J5NHaJhjh#0J5aJha4=0J5aJh=Bh=BnHo(tHh|nHtHhpnHtHhy*9nHtHhGnHtHh0h9haJmH sH hpOh9hCJ hpOCJhN`hN`CJ h>@KCJ7$78797P7U7l7m777777777788;8Y8k8l8m8o8y8}8888888889 9 99߳ߣ}umuuumccmmhNHnHtHh}\nHtHh]nHtHhBnHtHhnHtHha4=nHtHh9nHtH h 0Jhjh@5CJaJnHtHha4=5CJaJnHtHh}\5CJaJnHtH#hjh 5CJNHaJnHtHhjh 5CJaJnHtHhjh#5CJaJnHtH&999D9H9T9U9b9c999999999 ::#:%:&:4:5:6:?:A:B:H:V:X:n:y:{:::::::::޽𕋕xkxhBhj@ NHnHtHhBhj@ nHtHh~qnHtHhj@ NHnHtHhj@ nHtHh7Mhw*nHtHh7M6nHtHh7Mh7M6nHtHh7Mhw*6nHtHhw*6nHtHhw*hw*6nHtHhw*nHtHhNHnHtHhnHtHha4=nHtH(:::::::::;;;;;.;<;=;>;J;\;];_;b;j;m;;;;;;;;;···ʣʊtld[hw*6nHtHhw*nHtHhGWanHtHhkhBnHtHhr/FhBnHtHhr/Fhw*nHtHhkhw*6NHnHtHhkhw*6nHtHhknHtHhpOh~qnHtHhBnHtHhkhw*nHtHhBhw*nHtHhBhj@ NHnHtHhBhj@ nHtHhr/FnHtH ;;;;;;;;;;;<<<!<#<+<,<.<=<D<E<H<I<J<_<e<o<p<q<z<<<<<<<<<<<<<<<<<uhf6{NHnHtHh}nHtHh*nHtHhf6{nHtHhj@ nHtHhjnHtHha4=NHnHtHha4=nHtHhBnHtHhGonHtHjhr/F0JUnHtHhmcnHtHhw*hw*6nHtHhw*6nHtHhw*6NHnHtH.<<<<<<<<=&='=/=2=H=K=X=Y=q=r=========o[oD,h"hj@ OJPJQJ^JmH nH-sH tH-&hpOJPJQJ^JmH nH-sH tH-,h"h*OJPJQJ^JmH nH-sH tH- h"h*NHmH nH-sH tH-hpmH nH-sH tH-h"hj@ mH nH-sH tH-h"h*mH nH-sH tH-h"hj@ mH nHsH tHh}mH nHsH tHh omH nHsH tHh]nHtHhf6{nHtHha4=nHtH==>:>;>s>t>>>>>>>>>>??? ? ? ?)?-?6?7?CUCkC\QHH $Ifgd $G$Ifgd Mkd<$$If0 F 3F44 lal/$G$If`gd :kd$$IfFFF44 lal $If`gdACCCC"C#C*C+C,C8C>CJCUCeCkC}CCCCC񹩞nVnB'h^hr/F6CJaJmH nHsH tH.jhS^0J6CJUaJmH nHsH tH!hS^6CJaJmH nHsH tH$h^hr/FCJaJmH nHsH tHh^hr/F6CJaJh^hr/FCJaJh^hr/F6CJaJnHtHh^hr/FCJaJnHtH*h^hA56CJaJmH nHsH tH$hA56CJaJmH nHsH tHh^hACJaJnHtHCCCCCCCCCCCCCCCC DD"D%D,DGDIDJDرzrici]iSiOKD@h h9- hGh|hjh^hr/F6CJ hACJ h_LCJh^hr/FCJh^hr/F5h^hr/F]mH nHsH tHh^hr/F6CJaJh^hr/FCJaJh^hr/F6CJaJnHtHh^hr/FCJaJnHtH.jh0E0J6CJUaJmH nHsH tH'h^hr/F6CJaJmH nHsH tH%hS^6CJNHaJmH nHsH tHkCCCCCCC%DIDEGnI|wrmhgdTgdA gdYS?#gdG0gdr/FMkd$$If0 F 3F44 lal /$If`gd $Ifgd $G$Ifgd /$$If`a$gd JDLDYDaDbDqDrDwDxDDDDDDDDDD0E1EJEPETEZE^E_EvEEEEEEEEE5F6F7FDFKFQFUFeFFFо𚑚|u hAh hAhz hU=hU= hU=hushU=hzNH hU=hz hzNHhU=hAhU=mH nHsH tHhushAhA6hT hANHhAhp hkhz hkhUhk hUNHhzhUhh ,FFFFFFFFFFG(GtGGGGGGGGGGGGGGGGGG=H>HBHLH{HHHHHôxmbmbhy~hU=nHtHhy~hZ{nHtHh;qDnHtHhZ{NHnHtHhZ{nHtHh nHtHhpnHtHh hU=hZ{ hU=h;qDhTh;qDmH nHsH tH h Lh;qDh LhAhsNH hAhs hAhp%hA hAhU= hAhhAhNH&HHHHHHHHHHHHIII+I,I3I=I>IcIdImInIzIIIIIIIIII꿷vkvkvk`UhZ{hq;mH sH hZ{h mH sH hUh mH sH hUhq;mH sH hh mH sH  h9- hGhGnHtHhZ{NHnHtHhCNHnHtHhCnHtHhZ{nHtHha0nHtHhy~hZ{NHnHtHhy~hU=nHtHhy~h;qDnHtHhy~hZ{nHtHhy~hCnHtH!nIzIEKvLiMNOPQQ RRS5TT/UVVWWdXXxgdq3gd>@Kgd>@Kgd43xxgdW3gdL/gdL/gd8gdgdYgdzGgdqL~$gdGIIIIIIIIIIIIIIIJJJ}JJJJJJJJJJ*K+K/K0KķįģzoboboZhmH sH hqh NHmH sH hqh mH sH hkh mH sH hkhmH sH hmH sH hqL~h mH sH h7kh 6mH sH h mH sH h7kh NHmH sH h7kh mH sH hzGh mH sH hq;mH sH h mH sH hZ{hq;mH sH hZ{h mH sH 0K6K=K>KEKKKKKKKKK LeLvLLLLLLLLLL9M:MXMiMMʿҿzn`nUh+0hYmH sH h$h$6NHmH sH h$h$6mH sH h$mH sH hy=vh$mH sH hVmH sH hy=vh$mH sH hy=vh$6mH sH h$h$mH sH hy=vmH sH hy=vhy=vmH sH hy=vmH sH hy=vhy=vmH sH hqh NHmH sH hqh mH sH h4hmH sH MMMMM3NSNNNNNNNNNNOO#O$OOOOOOOOOOOOOPPP`PaPPPɻհzzobozozh4h8NHmH sH h4h8mH sH h8mH sH hk.h8mH sH  h NHhh 6NHhh 6 hh h h4hYmH sH h+0hY6NHmH sH h+0hY6mH sH hYmH sH h+0hYmH sH h+0hYmH sH h+0hYNHmH sH &PPQQQQ$Q%Q)Q2Q5QLQMQQQQQQQQ R RRRR#R&R*R+R0R3RARJRKRaRȻȻȳyk_h*okh*ok]mH-sH-h*okh*ok6]mH-sH-h*okh*okNHmH-sH-h*okmH-sH-h*okh*okmH-sH-h`hK 6NHmH-sH-h`hK 6mH-sH-hK mH-sH-hTiVhYNHmH sH hTiVhYmH sH hYmH sH hqhL/mH-sH-hL/hL/mH-sH-hC60hL/6 hC60hL/"aRbRfRhRRRRRRRRRRRRRRRRRRRRRRRɾ~l~`Phqh \mH-nH-sH-tH-h mH-nH-sH-tH-#hqh 6NHmH-nH-sH-tH-hqh 6mH-nH-sH-tH-hqh NHmH-sH-hQNmH-sH-h mH-sH-hq;mH-sH-hqh mH-sH-h*okh*okmH sH h*okmH-sH-h*okh*ok6mH-sH-h*okh*okmH-sH-h*ok]mH-sH-h*okh*okNH]mH-sH-RRS1S2SkSSSS"T#T5TOTVTWTcTdT{TTTTTTTTTTU$UxtpiaWaiP hqh h2&h 6NHh2&h 6 h2&h h hq;h4h>@KmH sH h>@Kh>@KmH sH hY^_h>@K6NH h>@K6hY^_h>@K6h>@Kh4h NHmH sH h4h mH sH h4h 6mH sH hq;h NHmH-sH-hq;h mH-sH-h mH-nH-sH-tH-hqh mH-nH-sH-tH-$U/U^UUUUUUUUVV5V;VvVwVVVVVVVVVV^W귯}qcq[PHhhmH sH hhhqmH sH hqmH sH h#hq6H*mH sH h#hq6mH sH h#hc6H*mH sH h#hc6mH sH hchcNHmH sH hchcmH sH hcmH sH hchcmH sH h*6H*mH sH h*6mH sH h#h*6mH sH h*mH sH h*h*mH sH hzGh mH sH ^WWWWWWWWWWWXX*X6X8X;X?XTXUX\XcXdXXXXXXXXYȽвЧwwod\d\d\dQhYhq;mH sH hYmH sH h`khYmH sH h mH sH hqL~h mH sH h7kh 6mH sH hqh 6mH sH hqL~h NHmH sH hqL~h mH sH h7kh mH sH hq;h mH sH hmH sH h mH sH hq;mH sH hqh mH sH hhmH sH hhhh6mH sH YY.Y/Y8Y:YFYPYVYnYtY|YYYYYYYYYYYZZ:Z;ZZZZZZеxkxcWh#h#6mH sH h LmH sH h Lh LNHmH sH h Lh LmH sH hfYNHmH sH hfYmH sH hfYhfYmH sH h mH sH h4h NHmH sH h4h mH sH hq;mH sH h mH sH h4h mH sH h4h 6NHmH sH h4h 6mH sH hYh mH sH XnYY;Z[[[[|\\6]Y]]]]]]]]]]]]]]gdGO$a$gdy*gdS^*$a$gd0E*$a$gd L**gd gd4ZZZ[[ [ [[[["[#[$[%[+[-[3[4[[[[[[[[[[[[[[[!\"\c\|\}\ǿǿǿǷǫǐǷ{tg{aZ h/>thg hgNHh|VhgOJQJ^J h>hg hahghgjhg0JUh4h NHmH sH h4h 6NHmH sH h4h 6mH sH h mH sH hq;mH sH h4h mH sH h Lh LmH sH h#mH sH h#h#6mH sH h#h#6NHmH sH #}\\\\] ]5]6]7]X]Y]Z][]^]u]v]]]]]]]]]]]]]]]]]]]]]]ǼǼǼǼǼh mH sH jh Uh h0EhgNHmH sH h0EhgmH sH hgmH sH hS^hgmH sH h C5hgmH sH  hgNHjhg0JU hr/Fhghg$]]gd46&P P:pGO. A!"#$% T= 0&P P:pGO. A!"#$% P TK$$If!vh5#v:V l t65T$$IfB!vh54 5E#v4 #vE:V FO54 5E/ 44 FaBTk$$If!vh5=5o5#v=#vo#v:V'l to 65Tk$$If!vh5=5o5#v=#vo#v:V'l to 65Tk$$If!vh5=5o5#v=#vo#v:V'l to 65Tk$$If!vh5=5o5#v=#vo#v:V'l to 65Tk$$If!vh5=5o5#v=#vo#v:V'l to 65TI$$If!vh5F#vF:V F5F/ al$$If!vh5 53#v #v3:V F5 53/ / / / al{$$If!vh5 53#v #v3:V F5 53/ / / al6666666666vvvvvvvvv666666>6666666666666666666666666666666666666666666666666hH6666666666666666666666666666666666666666666666666666666666666666668@8 GNormal_HmH sH tH\@b\ *okTtulo 6dd@&[$\$5CJPJ\aJmH sH tH NA@N Fuente de prrafo predeter.Ri@R  Tabla normal4 l4a ,k, Sin lista ROR GEACL Text Indent$`a$CJPJZOZ GEACL Text Indent CharCJPJ_HmH sH tHB&@B GRef. de nota al pieH*BO"B G EACL Address$a$CJPJ|OB| GEACL Abstract Heading'$ & F^`a$5CJPJaJXOBX G EACL Abstract$TT]T^Ta$CJPJLObL G EACL Title $,a$5CJPJ\aJDO"D G EACL Author$a$ 5CJPJJOrJ G EACL Email$<<a$ CJOJQJZOZ  EACL Reference textx^`CJHOH  EACL Reference text CharnOn GEACL Example 1st Line!!]!^!CJOJQJ^JaJPOP GEACL Example Last Line nOn GEACL List Bulleted#$ & FBx^B`a$ CJPJaJVOV GEACL List Bulleted 1st LinexJOJ GEACL Text CourierOJQJ^JaJ4O4 GEACL Text Char6O6 G EACL Text `NON "qUEACL book/journal title!6TO!T !qUEACL book/journal title Char6^O^  EACL Section#$$ & Fa$5CJPJ\aJjOj GEACL References Heading$$xa$5CJPJ\aJRO1R EACL Subsection% & FxCJaJDObD GEACL footnote text&CJ~@s~ GTabla con cuadrcula7:V'0'8O8 B EACL Example(2"@2 f0Epgrafe)5\8@8 >Texto nota pie*D'D #Ref. de comentarioCJaJ<< #Texto comentario,LjL #Asunto del comentario-5\LL #Texto de globo.CJOJQJ^JaJ^O^ (aFLREC main body text/$d$`a$PJtHPOP 1r/F LREC footnote0$a$CJPJmH sH tHNON 0r/FLREC footnote CarCJ_HmH sH tHTO"T (. LREC Address2$d@1$a$KHPJ nHtHtO2t L/LREC biblio references3$:d$^`:a$PJmH sH tHZ^@BZ *ok Normal (Web)4dd[$\$CJPJaJmH sH tH ,X@Q, *oknfasis6]r-38;;UA# U  8  gU@0@0@0()*=KgLY ' t aUWR- "x$&))G+P,----------------------L.M../m047E8F88889L9q999:;;;;#;+;,;>;U;k;;;;;;;%<I<=?nAzAECvDiEFGHII JJK5LL/MNNOOdPPnQQ;RSSSS|TT6UYUUUUUUUUUUUUUU0 002020202020200 0 0 0 0 0 0 0000 #0#00 /0 /0 /0 /0#0 #0#0#0#0 %0W0 0 0000000 %0W#0000/0 /0 /0 0 /0 /0 /0 0 /0 /0 /0 0 /0 /0 /0 0 /0 /0 /0 0 00 #0#000000000000000000 0 /0 /0 0 000/0 000/0 0 00 #0 000$000000000303030003000000000000@*0@*0@*0@*0@*0@*0@*00&0U@000@000@000@000@0@000 ()*=KgLY ' t aUWR- "x$&))G+P,----------------------L.M../m047E8F88889L9q999:;;;;#;+;,;>;U;k;;;;;;;%<I<=?nAzAECvDiEFGHII JJ5LL/MNNOOdPPnQQ;RSSS6UUUUU0 0020202020202000 0 0 00 0 0000 #0#00 /0 /0 /0 /0#0 #0#0#0#0 %0:0 0 0000000 %0:#0000/0/0/00/0/0/00/0/0/00/0/0/00/0/0/0000 #0#000000000000000000 0 /0 /0 0 000/0 000/0 0 00 #0 000$0000@0000I0s0K0s0@0@00300]00]0K0x0yv0L00N0I0\00Y0K0\00H000@0@0K00l 0&0UK00l 0&0U G & 2 r _\'%[~ >!7" $%&n()++M,J-R/00G1134679:;<=\?7@AA6BCCJDFHI0KMPaRR$U^WYZ}\]/3456789:;<=>?@BCDEFGHIJKLMNOPQRSTVWXYZ[\]cdefghiklmnoqrtuvxyz{|}~t*55555<CkCnIX]]02AU^_`abjpsw]18@0(  B S  ?3ZVZpZSZ:ZܬZpZ4ZZZaZDOZ|6ZDzZd>ZtZ<&ZDqZ@Z$ZcZIZ4qZKZ_ZZ̏[U[U[|rw[rw[[[4\[+[+[L [ [\ [ [  [[T[[[[< [| [ [-CCCClDsDsD_EfEfEFFFFFGGG IIIzIzIIIILLLM'M'MNNNNNOOPPPQRRRSU  !"#$%&'()*+,-.0/12>*urn:schemas-microsoft-com:office:smarttags PersonNameC.*urn:schemas-microsoft-com:office:smarttagsmetricconverter #urn:schemas:contactsSn93*urn:schemas-microsoft-com:office:smarttagsplace=*urn:schemas-microsoft-com:office:smarttags PlaceName=*urn:schemas-microsoft-com:office:smarttags PlaceTypeB1*urn:schemas-microsoft-com:office:smarttagscountry-region82*urn:schemas-microsoft-com:office:smarttagsCity  1 a1995. Ala Universitat ProductID32132.3123123123#3##3#231233333223123#3#.321!'*12;evIY 0`c#)#&, ? O   + 3 ~ PSTXem <A/7<CGW+1LU f!h!i!l!v!y!}!!!!!!!!""9$F$G$K$%%''''(((((())X*h*,,,---------------....//00[0d0/525H5K5556666)7,7<7@7]888888888888899R9T9p99999:/:l:|:::::;;;;;;;;;;;;;<<"<j>z>@ @@@zAAAAAAAAAAAAAAAAAAAAAABBBBBBBBBBBBCC)C/CECKCPCYC^ChCmCvC{CCCCCCCCCCDDvD|DDDDDDDDDDDiErEwEEEEEEEEEEEEEEEEEEF FFFF!F"F+F2F8FDFJFFFG!G"G+G1G8GCGJGYG`GaGiGGGGGGGHHI IKIZIIIIIIIIIIIIIJ JJJJ#J(J/J7JAJVJYJ^JfJhJsJtJvJJJJJJJJJJJJJJJJJJJKKKK KkKoKpKwKxK{KKKKKKKKKKKKKKKLL LLL(L)L3LFLMLLL(M-M/M7MCMJMRMZMMMNN!N$N)N,N-N1NsN{N}NNNNNNOOqO{OOOOOOO?PHPIPTPdPlPqPwPPPPPPQ:QFQHQRQSQUQtQzQQQ,R/R?RBRCRGRLRURRRRRRR]SeSuS}SSSSSSSST/TdTmTTTUUUUUUUUUUUUUU ''55F888888899K9L9p9q9y999;#<1C4CCD DfDDDDYEEESFFFFFG-GGGGGGGGGGGHHHHHHIIJIKIsItIIIIIIIJ JJJJKkKK4L5LEL}LLLM.M/McMdMMMNNN O&O^OOOOOO8P]PdPPQQQQ;RRRR]SuSSUUUUUUUUUUUUUU333333333333333333333333333333333333333333333333=K*1LY))----------16l6F8888q9;>;;;DDFGGGIIJJ5LFL}LL^PcPnQQQQSUUUUUUUUUUUUUUSUUUUUUUUUUUUUUC\-bN AJYIJvd#% kRX@ W:`hX-jb6^`. P^`Po(hH @@^@`o(hH. 0^`0o(hH.. ``^``o(hH... ^`o(hH .... ^`o(hH ..... ^`o(hH ......  `^``o(hH.......  00^0`o(hH........h^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hH P^`Po(hH ^`o(hH. 0^`0o(hH.. ``^``o(hH... ^`o(hH .... ^`o(hH ..... ^`o(hH ......  `^``o(hH.......  00^0`o(hH........hKK^K`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHoh[[^[`OJ QJ o(hHh++^+`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hH^`CJOJQJo(^`CJOJQJo(opp^p`CJOJ QJ o(@ @ ^@ `CJOJ QJ o(^`CJOJ QJ o(^`CJOJ QJ o(^`CJOJ QJ o(^`CJOJ QJ o(PP^P`CJOJ QJ o(ZZ^Z`OJPJQJ^Jo(-^`OJQJ^Jo(hHo^`OJ QJ o(hH  ^ `OJQJo(hH  ^ `OJQJ^Jo(hHo[[^[`OJ QJ o(hH++^+`OJQJo(hH^`OJQJ^Jo(hHo^`OJ QJ o(hHC\-IJAJIJIJ-jIJIJIJIJIJIJIJIJ kRIJIJIJIJIJIJIJIJIJIJIJIJ W:`WW8Num25H        ---------:-------- ~ 9)' x3HvWb9)'Hi\=,x3V K k8<6P[HV K k8<6P~ vWb[H2 1B_L`{o~P!NE0:+Go,vPY; K "^  ! h j@ F 9- 3 M o oL@EnY0=uv6Qx> q^ w*1Hp6AhGiY%k !hw pp%D0E-.P i s { \!#p#Jr#;<%E'Ds'2(p(o)i *A*!+7$+J+ ,O,--i-(o-K./b/c/a001M2<2TO3{~344}S4&5 C5'6z6Q799y*9cH9[9=;q;"<a4=U= >->J>c>"B?YS?Z[?c?k? @q B]BB=BrnC/DKD;qD-Er/F(aFVwF xG{G5H^9H.JHuHI[I{eIzIwKK>@KaK LMdM?NjOpOyO%zOXRjRXSK|SnTn[TqU\VQVTiVsXYY1iYZz%Z}}Z4[O5[~{[Ki\}\^S^*`UL`N`8t`J~`aa(a6aGWacmcjd9-d e e5)g9h:h;Oi`ai7jj|j9#jlj7kgQk`k*ok_lyl1m3n opp?pqq{"qbr(rus?sFs8Ls/>t]tftu*0uvT vy=vgvmvk{w@SxDyWyiywyPz{@*{+{f6{O{Y{a{|||]}P~qL~y~J!N ar "!*Z.3d7yZ{g-HM|}@'pQ:DIt-;9=oXs-G, p09/GI= BzGGF_b0QN :_u jxA|Vbagz2&"Eo]t qzIP2Y=-s:M/fvos` Huy13!CS[ UqypzV,-FjlvQt FYF +0[| 9%I4`gYrGwg!5bf/hJe 7M$_yqrzaD U&? R.hT1i9v]d YmLW}**3Gt`\z0@`Rk}%B]P>cV1ZE\#<25SD-J)24egnP7w8fY}__vGVb 3/4PZ'!cmp!38 Wj-lXq~qL/(.qQ]+Rbnk.wGGOsz +b~f7=Y>0GcloBR C+ecN$v/0BeW [}a~!#>2^NTB?@ w*LU ----------------------q9;;;#;+;,;;;;?CISS6UU-E}\E0 @ YYPoYYU@Unknown Gz Times New Roman5Symbol3& z Arial?5 z Courier NewCTimesNewRomanQTimesNewRoman,ItalicC PMingLiUe0}fԚG  MS Mincho-3 fg5& zaTahoma;SimSun[SO;Wingdings"1eff F* F*4dqSqS 2qHX 0G2%Instructions for ACL-2010 ProceedingsJing-Shin Chang & Philipp Koehnacpalroi(       Oh+'0, @L l x  (Instructions for ACL-2010 Proceedings Jing-Shin Chang & Philipp KoehnNormal acpalroi2Microsoft Office Word@@$o[@:#_@:#_ F՜.+,0( hp  ACL-2010 Publication Chairs*qS' &Instructions for ACL-2010 Proceedings Ttulo  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry Fk+9_Data 1Table|WordDocumentw SummaryInformation(DocumentSummaryInformation8CompObjr  F Documento Microsoft Office Word MSWordDocWord.Document.89q