File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/93/p93-1021_metho.xml
Size: 19,054 bytes
Last Modified: 2025-10-06 14:13:30
<?xml version="1.0" standalone="yes"?> <Paper uid="P93-1021"> <Title>A LANGUAGE-INDEPENDENT ANAPHORA RES()LUTION SYSTEM FOR UNDERSTANDING MULTILINGUAL TEXTS</Title> <Section position="1" start_page="0" end_page="0" type="metho"> <SectionTitle> A LANGUAGE-INDEPENDENT ANAPHORA RES()LUTION SYSTEM FOR UNDERSTANDING MULTILINGUAL TEXTS </SectionTitle> <Paragraph position="0"/> </Section> <Section position="2" start_page="0" end_page="0" type="metho"> <SectionTitle> Abstract </SectionTitle> <Paragraph position="0"> This paper describes a new discourse module within our multilingual NLP system. Because of its unique data-driven architecture, the discourse module is language-independent. Moreover, the use of hierarchically organized multiple knowledge sources makes the module robust and trainable using discourse-tagged corpora. Separating discourse phenomena from knowledge sources makes the discourse module easily extensible to additional phenomena.</Paragraph> </Section> <Section position="3" start_page="0" end_page="0" type="metho"> <SectionTitle> 1 Introduction </SectionTitle> <Paragraph position="0"> This paper describes a new discourse module within our multilingual natural language processing system which has been used for understanding texts in English, Spanish and Japanese (el. \[1, 2\])) The following design principles underlie the discourse module: * Language-independence: No processing code depends on language-dependent facts.</Paragraph> <Paragraph position="1"> * Extensibility: It is easy to handle additional phenomena. null * Robustness: The discourse module does its best even when its input is incomplete or wrong.</Paragraph> <Paragraph position="2"> * Trainability: The performance can be tuned for particular domains and applications.</Paragraph> <Paragraph position="3"> In the following, we first describe the architecture of the discourse module. Then, we discuss how its performance is evaluated and trained using discourse-tagged corpora. Finally, we compare our approach to other research.</Paragraph> <Paragraph position="4"> 1 Our system has been used in several data extraction tasks and a prototype nlachine translation systeln.</Paragraph> <Paragraph position="5"> perfo.m .... ~nti ~u2k c$~ &quot; e dv ....... ...................</Paragraph> <Paragraph position="6"> r ............ . ....... o ..................................... -, l:)i~ ~ Module</Paragraph> </Section> <Section position="4" start_page="0" end_page="159" type="metho"> <SectionTitle> 2 Discourse Architecture </SectionTitle> <Paragraph position="0"> Our discourse module consists of two discourse processing submodules (the Discourse A dministralor and the Resolution Engine), and three discourse knowledge bases (the Discourse Knowledge Source KB, the Discourse Phenomenon KB, and the Discourse Domain KB). The Discourse Administrator is a development-time tool for defining the three discourse KB's. The Resolution Engine, on the other hand, is the run-time processing module which actually performs anaphora resolution using these discourse KB's.</Paragraph> <Paragraph position="1"> The Resolution Engine also has access to an external discourse data structure called the global discourse world, which is created by the top-level text processing controller. The global discourse world holds syntactic, semantic, rhetorical, and other information about the input text derived by other parts of the system. The architecture is shown in Figure i.</Paragraph> <Section position="1" start_page="0" end_page="156" type="sub_section"> <SectionTitle> 2.1 Discourse Data Structures </SectionTitle> <Paragraph position="0"> There are four major discourse data types within the global discourse world: Discourse World (DW), \[)is null course Clause (DC), Discourse Marker (DM), and File Card (FC), as shown in Figure 2.</Paragraph> <Paragraph position="1"> The global discourse world corresponds to an entire text, and its sub-discourse worlds correspond to sub-components of the text such as paragraphs. Discourse worlds form a tree representing a text's structure. A discourse clause is created for each syntactic structure of category S by the semantics module. It can correspond to either a full sentence or a part of a flfll sentence. Each discourse clause is typed according to its syntactic properties.</Paragraph> <Paragraph position="2"> A discourse marker (cf. Kamp \[14\], or &quot;discourse entity&quot; in Ayuso \[3\]) is created for each noun or verb in the input sentence during semantic interpietation.</Paragraph> <Paragraph position="3"> A discourse marker is static in that once it is introduced to the discourse world, the information within it is never changed.</Paragraph> <Paragraph position="4"> Unlike a discourse marker, a file card (cf. Heim \[11\], &quot;discourse referent&quot; in Karttunen \[15\], or &quot;discourse entity&quot; in Webber \[19\]) is dynamic in a sense that it is continually updated as the discourse processing proceeds. While an indefinite discourse marker starts a file card, a definite discourse marker updates an already existing file card corresponding to its antecedent. In this way, a file card keeps track of all its co-referring discourse markers, and accumulates semantic information within them.</Paragraph> </Section> <Section position="2" start_page="156" end_page="157" type="sub_section"> <SectionTitle> 2.2 Discourse Administrator </SectionTitle> <Paragraph position="0"> Our discourse module is customized at development time by creating and modifying the three discourse KB's using the Discourse Administrator. First, a discourse domain is established for a particular NLP application. Next, a set of discourse phenomena which should be handled within that domain by the discourse module is chosen (e.g. definite NP, 3rd per-son pronoun, etc.) because some phenomena may not be necessary to handle for a particular application domain. Then, for each selected discourse phenomenon, a set of discourse knowledge sources are chosen which are applied during anaphora resolution, since different discourse phenomena require different sets of knowledge sources.</Paragraph> <Paragraph position="1"> The discourse knowledge source KB houses small well-defined anaphora resolution strategies. Each knowledge source (KS) is an object in the hierarchically organized KB, and information in a specific KS can be inherited from a more general KS.</Paragraph> <Paragraph position="2"> There are three kinds of KS's: a generator, a filter and an orderer. A generator is used to generate pos-</Paragraph> <Paragraph position="4"> sible antecedent hypotheses from the global discourse world. Unlike other discourse systems, we have multiple generators because different discourse phenomena exhibit different antecedent distribution patterns (cf.</Paragraph> <Paragraph position="5"> Guindon el al. \[10\]). A filter is used to eliminate impossible hypotheses, while an orderer is used to rank possible hypotheses in a preference order. The KS tree is shown in Figure 3.</Paragraph> <Paragraph position="6"> Each KS contains three slots: ks-flmction, ks-data, and ks-language. The ks-function slot contains a functional definition of the KS. For example, the functional definition of the Syntactic-Gender filter defines when the syntactic gender of an anaphor is compatible with that of an antecedent hypothesis. A ks-data slot contains data used by ks-function. The separation of data from function is desirable because a parent KS can specify ks-function while its sub-KS's inherit the same ks-function but specify their own data. For example, in languages like English and Japanese, the syntactic gender of a pronoun imposes a semantic gender restriction on its antecedent. An English pronoun &quot;he&quot;, for instance, can never refer to an NP whose semantic gender is female like &quot;Ms.</Paragraph> <Paragraph position="7"> Smith&quot;. The top-level Semantic-Gender KS, then, defines only ks-flmction, while its sub-KS's for English and Japanese specify their own ks-data and inherit the same ks-function. A ks-language slot specifies languages if a particular KS is applicable for specific languages.</Paragraph> <Paragraph position="8"> Most of the KS's are language-independent (e.g.</Paragraph> <Paragraph position="9"> all the generators and the semantic type filters), and even when they are language-specific, the function ; DW date of the text ; loc~tion where the text is originated ; semantic concepts which correspond to globM topics of the text ; the corresponding character position in the text ; ~ list of discourse clauses in the current DW ; a list of DWs subordinate to the current one (defframe discourse-clause (discourse-d~ta-structure ; D(: discourse-markers ; ~ list of discourse m~rkers in the current D(:~ syntax ; ~n f-structure for the current DC parse-tree ; ~ p~rse tree of this S semantics ; ~ semantic (KB) object representing the current DC position ; the corresponding character position in the text d~te ; date of the current DC~ loca.tion ; Ioco.tlon of the current D(2 subordinate-discourse-clsuse ; a DC,&quot; subordinate to the current D(: coordin~te-dlscourse-clattses) ; coordinate DC's which a conjoined sentence consists of II (dell di .... ........... ker(dl d ture' ;DM ........ ......... Jr position ; the corresponding character position in the text discourse-clause ; a pointer b~ck to DC: syntax ; an f-structure for the current DM semantics ; a semantic (KB) object file card) ; a pointer to the file card definitions are shared. In this way, much of the discourse knowledge source KB is sharable across different languages.</Paragraph> <Paragraph position="10"> The discourse phenomenon KB contains hierarchically organized discourse phenomenon objects as shown in Figure 4. Each discourse phenomenon object has four slots (alp-definition, alp-main-strategy, dp-backup-strategy, and dp-language) whose values can be inherited. The dp-definilion of a discourse phenomenon object specifies a definition of the discourse phenomenon so that an anaphoric discourse marker can be classified as one of the discourse phenomena. The dp-main-strategy slot specifies, for each phenomenon, a set of KS's to apply to resolve this particular discourse phenomenon. The alp-backupstrategy slot, on the other hand, provides a set of backup strategies to use in case the main strategy fails to propose any antecedent hypothesis. The dp-language slot specifies languages when the discourse phenomenon is only applicable to certain languages (e.g. Japanese &quot;dou&quot; ellipsis).</Paragraph> <Paragraph position="11"> When different languages use different sets of KS's for main strategies or backup strategies for the same discourse phenomenon, language specific dp-main-strategy or dp-backup-strategy values are specified. For example, when an anaphor is a 3rd person pronoun in a partitive construction (i.e. 3PRO-Partitive-Parent) 2, Japanese uses a different generator for the main strategy (Current-and-Previous-DC) than English and Spanish (Current-and-Previous-Sentence).</Paragraph> <Paragraph position="12"> 2e.g. &quot;three of them&quot; ill English, &quot;tres de ellos&quot; in Spanish, &quot;uchi san-nin&quot; in Japaamse Because the discourse KS's are independent of discourse phenomena, the same discourse KS can be shared by different discourse phenomena. For example, the Semantic-Superclass filter is used by both Definite-NP and Pronoun, and the Recency orderer is used by most discourse phenomena.</Paragraph> <Paragraph position="13"> The discourse domain KB contains discourse domain objects each of which defines a set of discourse phenomena to handle \[n a particular domain. Since texts in different domains exhibit different sets of discourse phenomena, and since different applications even within the same domain may not have to handle the same set of discourse phenomena, the discourse domain KB is a way to customize and constrain the workload of the discourse module.</Paragraph> </Section> <Section position="3" start_page="157" end_page="158" type="sub_section"> <SectionTitle> 2.3 Resolution Engine </SectionTitle> <Paragraph position="0"> The Resolution Engine is the run-time processing module which finds the best antecedent hypothesis for a given anaphor by using data in both the global discourse world and the discourse KB's. The Resolution Engine's basic operations are shown in Figure 5. The Resolution Engine uses the discourse phenomenon KB to classify an anaphor as one of the discourse phenomena (using dp-definition values) and to determine a set of KS's to apply to the anaphor (using dp-main-strategy values). The Engine then applies the generator KS to get an initial set of hypotheses and removes those that do not pass tile filter</Paragraph> <Paragraph position="2"/> </Section> <Section position="4" start_page="158" end_page="159" type="sub_section"> <SectionTitle> Find-Antecedent </SectionTitle> <Paragraph position="0"> Input: aalaphor to resolve, global discourse world Get-KSs-for-Discourse-Phenomenon Input: anaphor to resolve, discourse phenomenon KB Output: a set of discourse KS's Apply-KSs hlput: aalaphor to resolve, global discourse world, discourse KS's Output: the best hypothesis Output: the best hypothesis Update-Discourse-World Input: anaphor, best hypothesis, global discourse world Output: updated global discourse world the anaphor's referent, but there may be more than one hypothesis or none at all.</Paragraph> <Paragraph position="1"> When there is more than one hypothesis, orderer KS's are invoked. However, when more than one orderer KS could apply to the anaphor, we face the problem of how to combine the preference values returned by these multiple orderers. Some anaphora resolution systems (cf. Carbonell and Brown \[6\], l~ich and LuperFoy \[16\], Rimon el al. \[17\]) assign scores to antecedent hypotheses, and the hypotheses are ranked according to their scores. Deciding the scores output by the orderers as well as the way the scores are combined requires more research with larger data. In our current system, therefore, when there are multiple hypotheses left, the most &quot;promising&quot; orderer is chosen for each discourse phenomenon. In Section 3, we discuss how we choose such an orderer for each discourse phenomenon by using statistical preference. In the future, we will experiment with ways for each orderer to assign &quot;meaningful&quot; scores to hypotheses. When there is no hypothesis left after the main strategy for a discourse phenomenon is performed, a series of backup strategies specified in the discourse phenomenon KB are invoked. Like the main strutegy, a backup strategy specifies which generators, filters, and orderers to use. For example, a backup strategy may choose a new generator which generates more hypotheses, or it may turn off some of the filters used by the main strategy to accept previously rejected hypotheses. How to choose a new generator or how to use only a subset of filters can be determined by training the discourse module on a corpus tagged with discourse relations, which is discussed in Section 3.</Paragraph> <Paragraph position="2"> Thus, for example, in order to resolve a 3rd per-son pronoun in a partitive in an appositive (e.g. anaphor ID=1023 in Figure 7), the phenomenon KB specifies the following main strategy for Japanese:</Paragraph> <Paragraph position="4"> cency. This particular generator is chosen because in almost every example in 50 Japanese texts, this type of anaphora has its antecedent in its head NP. No syntactic filters are used because the anaphor has no useful syntactic information. As a backup strategy, a new generator, Adjacent-NP, is chosen in case the parse fails to create an appositive relation between the antecedent NP ID=1022 and the anaphor.</Paragraph> <Paragraph position="5"> After each anaphor resolution, the global discourse world is updated as it would be in File Change Semantics (cf. Helm \[11\]), and as shown in Figure 6. First, the discourse marker for the anaphor is incorporated into the file card to which its antecedent discourse marker points so that the co-referring discourse markers point to the same file card. Then, the semantics information of the file card is updated so that it reflects the union of the information from all the co-referring discourse markers. In this way, a file card accumulates more information as the discourse processing proceeds.</Paragraph> <Paragraph position="6"> The motivation for having both discourse markers and file cards is to make the discourse processing a monotonic operation. Thus, the discourse processing does not replace an anaphoric discourse marker with its antecedent discourse marker, but only creates or updates file cards. This is both theoretically and computationally advantageous because the discourse processing can be redone by just retracting the file cards and reusing the same discourse markers.</Paragraph> </Section> <Section position="5" start_page="159" end_page="159" type="sub_section"> <SectionTitle> 2.4 Advantages of Our Approach </SectionTitle> <Paragraph position="0"> Now that we have described the discourse module in detail, we summarize its unique advantages. First, it is the only working language-independent discourse system we are aware of. By &quot;language-independent,&quot; we mean that the discourse module can be used for different languages if discourse knowledge is added for a new language.</Paragraph> <Paragraph position="1"> Second, since the anaphora resolution algorithm is not hard-coded in the Resolution Engine, but is kept in the discourse KB's, the discourse module is extensible to a new discourse phenomenon by choosing existing discourse KS's or adding new discourse KS's which the new phenomenon requires.</Paragraph> <Paragraph position="2"> Making the discourse module robust is another important goal especially when dealing with real-world input, since by the time the input is processed and passed to the discourse module, the syntactic or semantic information of the input is often not as accurate as one would hope. The discourse module must be able to deal with partial information to make a decision. By dividing such decision-making into multiple discourse KS's and by letting just the applicable KS's fire, our discourse module handles partial information robustly.</Paragraph> <Paragraph position="3"> Robustness of the discourse module is also manifested when the imperfect discourse KB's or an inaccurate input cause initial anaphor resolution to fail. When the main strategy fails, a set of backup strategies specified in the discourse phenomenon KB provides alternative ways to get the best antecedent hypothesis. Thus, the system tolerates its own insufficiency in the discourse KB's as well as degraded input in a robust fashion.</Paragraph> </Section> </Section> class="xml-element"></Paper>