File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/00/w00-1414_metho.xml
Size: 29,344 bytes
Last Modified: 2025-10-06 14:07:25
<?xml version="1.0" standalone="yes"?> <Paper uid="W00-1414"> <Title>Generating Referring Quantified Expressions</Title> <Section position="5" start_page="100" end_page="100" type="metho"> <SectionTitle> 3 The Application Domain </SectionTitle> <Paragraph position="0"> We implemented our quantification algorithm as part of MAGIC (Dalai et al., 1996: McKeown et al., 1997). MAGIC automatically generates multi-media briefings to describe the post-operative status of a patient after undergoing Coronary Artery Bypass Graft, surgery. The system embodies a stan- null &quot;After intubation, a patient received aprotinin.&quot; dard text generation system architecture with three modules (Rambow and Korelsky, 1992): a content planner, a sentence planner, and a linguistic realizer. Once the bypass surgery is finished, information that is automatically collected during surgery such as blood pressure, heart rate, and medications given, is sent to a domain=specific medical inference module. Based on the medical inferences and schemas (McKeown, 1985), the content planner determines the information to convey and the order to convey it.</Paragraph> <Paragraph position="1"> The sentence planner takes a set of propositions (or predicate-argument structures) with rhetorical relations from the content planner and uses linguistic information to make decisions about how to convey the propositions fluently. Each proposition is represented as a feature structure (Kaplan and Bresnan, 1982; Kay, 1979) similar to the one shown in Figure 1. The sentence planner's responsibilities include referring expression generation, clause aggregation, and lexical choice (Wanner and How, 1996). Then the aggregated predicate-argument structure is sent to FUF/SURGE (Elhadad and Robin, 1992), a linguistic realizer which t.ransforms the lexicalized seinantic specification into a string. The quantification algorithm is implemented in the sentence planner.</Paragraph> </Section> <Section position="6" start_page="100" end_page="104" type="metho"> <SectionTitle> 4 Quantification Algorithm </SectionTitle> <Paragraph position="0"> in this:,work, weprefergenerating expressions with universal quantifiers over conjunction because, assuming that the users and the system have tile same domain model, the universally quantified expressions are more concise and they represent the same amount of information as the expression with conjoined entities. In contrast,, when given a conjunction of entities and an expression with a cardinal quantifier, the system, by default, would use the conjunction if the conjoined entities can be distinguished at the surface level. This is because once the system generates a cardinal quantifier when the universal quantification does not hold, such as &quot;three patients&quot;, it is impossible for the hearer to recover the identities of these patients based on the context. The default heuristics to prefer universal quantifier over conjunction over cardinal quantifier can be superseded by directives fromthe contentplanner which are application specific.</Paragraph> <Paragraph position="1"> The input to our quaatifica~omalgorit;hm is a set of predicate-argument structures after the referring expression module selected the properties to identify the entities (Dale, 1992; Dale and Reiter, 1995), but without carrying out the assignment of quantifiers.</Paragraph> <Paragraph position="2"> Our quantification algorithm first identifies the set of distinct entities which can be quantified in the input propositions. A generalization of the entities in the ontology is selected to potentially replace the references to these entities. If universal quantification is possible, then the replacement is made and the system must select which particular quantifier to use. In our system, we have six realizations for universal quantifiers: each, every, all 1, both, the, and any, and two for existential quantifiers: the indefinite article, a/an, and cardinal n.</Paragraph> <Section position="1" start_page="101" end_page="101" type="sub_section"> <SectionTitle> 4.1 Identify Thematic Roles with Distinct Entities </SectionTitle> <Paragraph position="0"> Our algorithm identifies the roles containing distinct entities among the input propositions as candidates for universal and existential quantification. Suppose the system is given two propositions similar to the one in Figure 1, &quot;After intubation, Alice received aprotinin&quot; and &quot;After start of bypass, Alice received aprotinin&quot;, each with four roles - PRED, ARG1, ARG2, and MODS-TIME. By computing similarity anaong entities in the same role, the system determines that the entities in ARG1, PRED, and ARG2 are identical in each role, and only the entities in MODS-TIME are different. Based on this result, the distinct entities in MODS-TIME, &quot;after intubation&quot; and &quot;after start of bypass&quot;, are candidates for quantificat ion.</Paragraph> </Section> <Section position="2" start_page="101" end_page="102" type="sub_section"> <SectionTitle> 4.2 Generalization and Quantification </SectionTitle> <Paragraph position="0"> We used the axioms in Figure 2 to determine if the distinct entities can be universally or existentially quantified. Though the axioms are similar to those used in Generalized Quantifier (Barwise and Cooper, 1981; Zwarts, 1983; de Swart, 1998). the semantics of set X and set D are different. In the previous step. the entities in set X have been identified. To compute set D in Figure 2. we introduce a concept, Class-X. Class-X is a generalization of the distinct entities in set X. Quantification can replace the distinct entities in the propositions with a reference to their type restricled by a quantifier.</Paragraph> <Paragraph position="1"> accessing discourse and ontological information .to provide a context. Our ontology is implemented in lali is realized as &quot;ali the&quot;.</Paragraph> <Paragraph position="2"> * both: ID - X\[ = 0 and IxI = 2, can have collective reading * every, all, the: ID-XI = 0 and IX\[ > 2, can have collective reading * each: I D - X\[ = 0 and IXI _> 2, only distribu-- * tive reading ........</Paragraph> <Paragraph position="3"> (r) any: \]D- X\] = 0, when under the scope of negation deg a/an: IDnXl > 0 and Ixl = 1 * n (cardinal): IOnXl > 0 and \[Xl = n CLASSIC(Borgida et al., 1989) and is a subset of WordNet(Miller et alL, 1990) and an online medical dictionary (Cimino et al., 1994) designed to support multiple applications across the medical institution. Given the entities in set X, queries in CLASSIC determine the class of each instance and its ancestors in the ontology. Based on this information, the generalization algorithm identifies Class-X by computing the most specific class which covers all the entities. Earlier work (Passonneau et al., 1996) provided a framework for balancing specificity and verbosity in selecting appropriate concepts for generalization. However, given the precision needed in medical reports, our generalization procedure selects the most specific class.</Paragraph> <Paragraph position="4"> Set D represents the set of instances of Class-X in a context. Our system currently computes set D for three different contexts: e discourse: Previous references can provide an appropriate context for universal quantification. For example, if &quot;Alice&quot; and &quot;Bob&quot; were mentioned in the previous sentence, the system can refer t.o them as &quot;both patients&quot; in the current sentence.</Paragraph> <Paragraph position="5"> (r) domain ontology: The domain ontology provides a closed world from which we can obtain 't-he set D by matching all the instances of a concept in the knowledge base, such as &quot;'every patient&quot;. In addition, certain concepts in the ontology have limited types. For example, knowing that cell savers, platelets and packed red blood cells are the only possible types of blood products in the ontology, the quantified expression &quot;every blood product&quot; can be used instead of referring to each entity.</Paragraph> <Paragraph position="6"> (r) domain knowledge: The possessor of the distinct entities in a role might contain a maximum number of instances allowed for Class-X. For example, because a person has only two arms, the tinguishable expressions at surface level. A more entities &quot;the patient's left arm&quot; and &quot;the pa- developed pragmatic module is needed before quantient's right arm&quot; can be referred to as &quot;each tifiers such as some, raps'e, at least, and few, can arm&quot;. be systematically generated. Indiscriminate application of imprecise quantification can result in- vague The computation of set D can also involve interac- or inappropriate text in our domain, such as &quot;The tions with a referring expression m0dule(Dale aad~ ~-:patient~rec~ived.~some 61ood~produetS:&quot;'-v.ia-our~e~P -Reiter, 1995). For example, instead of the expres- plication, knowing exactly what blood products are sion &quot;Alice and Bob&quot; and &quot;both patients&quot; covered by the current algorithm, by interacting with a referring expression module, the system might determine that &quot;both CABG patients operated on this morning by Dr. Rose&quot; is a clearer expression to refer to the entities. Though this is desirable, we did not incorporate this capability into our system.</Paragraph> <Paragraph position="7"> Although the is often used to indicate a generic reference (i.e., &quot;The lion is the king of jungle.&quot;), in English, the can also be used as an unmarked universal quantifier when its head noun is plural, such as &quot;the patients.&quot; Like the quantifier all, the can be both distributive and collective. However, the cannot always replace all as a universal quantifier.</Paragraph> <Paragraph position="8"> the cannot be used when universal quantification is based on the domain ontology. For example, it is not obvious that the quantified expression in &quot;John received the blood products.&quot; refers to &quot;each blood product&quot; in the ontology. Although unmarked universal quantifiers can be used to refer to body parts, as in &quot;The lines include an IV in the arms.&quot;, the expression is ambiguous between the distributive and collective readings. Of the three contexts discussed above, the system occationally generates the instead of every and both in a discourse context, yielding more natural output.</Paragraph> <Paragraph position="9"> When the computed set D matches set X exactly</Paragraph> <Paragraph position="11"> each, all, every, both, the, and any, replaces the entities in set X.</Paragraph> </Section> <Section position="3" start_page="102" end_page="103" type="sub_section"> <SectionTitle> 4.3 Selecting a Particular Quantifier </SectionTitle> <Paragraph position="0"> In general, the universal quantification of a particular type of entity, such as &quot;every patient&quot;, refers to all such entities in a context. As a result, readers can recover what a universally quantified expression refers to. In contrast, readers cannot pinpoint which entity has been refei'red to. in an existentially .</Paragraph> <Paragraph position="1"> quantified expression, such as &quot;a patient.&quot; or &quot;two patients&quot;. Because a universally quantified expression preserves original semantics and is more concise than listing each entity, it is the focus of our quantificalion algorithm. The universal quantifiers hlaplemented in our system include the six possible realizations of V in English: every, all. each. both.</Paragraph> <Paragraph position="2"> the, and any. The only existential quantifiers implemented in our system are the singular indefinite quantifier, a/an. and cardinal quantifiers, n. They are used in sentences with multiple quantifiers and when the entities being referred to do not have disused is very important. To avoid generating such inappropriate sentences, the system only performs generalization on the entities which can be universally quantified. If the distinct entities cannot be universally quantified, the system will realize these entities using coordinated conjunction.</Paragraph> <Paragraph position="3"> Once the system decides that a universally quantified expression can be used to replace the entities in set X, it must select which universal quantifier.</Paragraph> <Paragraph position="4"> Because our sentence planner opportunistically combines distinct entries from separate database entries for conciseness, it is not the case that these aggregated entities acted together (the collective reading). Given such input, the referring expression for aggregated entities should have only the distributive reading 2. The universal quantifier, each, always imposes a distributive reading when applied.</Paragraph> <Paragraph position="5"> In general, each requires a &quot;matching&quot; between the domain of the quantifier and the objects referred to(McCawley, 1981, pp. 37). In our algorithm, this matching process is exactly what happened, thus it is the default universal quantifier in our algorithm.</Paragraph> <Paragraph position="6"> Of course, indiscriminate use of each can result in awkward sounding text. For example, tile sentence &quot;Every patient is awake&quot; sounds more natural than &quot;Each patient is awake.&quot; However, since quantified expressions with the universal quantifiers all and every 3 can have collective readings (Vendler, 1967; McCawley, 1981), our system generates every and all under two conditions when the collective reading is unlikely. First if the proposition is a state, as opposed to an event, we assume only the distributive reading is possible 4. The quantifier every is used in &quot;Ever.q patient tmd.taehycardia.'&quot; because the proposition is a state proposition and contains the predicate has-attribute, an attributive relation.</Paragraph> <Paragraph position="7"> ..... 2For our system to generate noun-phrases.wivh ,col}eetive readings, the quantification process must be performed at the content planner level not in the clause aggregation module. 3every is also distributive, but it stresses completeness or rather, exhaustiveness(Vendler, 1967). The sentence &quot;John took a picture of everyone in the room.&quot; is ambiguous while &quot;John took a picture os t each person in the room.&quot; is not. 4There are cases where state propositions do have disteibuted readings (e.g., &quot;Mountains surround the village.&quot; ). Sentences with collective readings are bandied earlier in the content planner and thus, this type of problem does not occur at this point in our system. Though .this observation seems to be true in our medical application, when implementing quantifiers in a new domain, we can limit this assumption to only the subset of state relations for which it holds.</Paragraph> <Paragraph position="8"> Second, when the concept being universally quantified is marked as having a distributive reading in the lexicon, such as the concept episode, quantifiers every will be used instead of each. These quantitiers make the quantified sentences more natural because they do not pick out the redundant distributive meaning. ...... .~ . -: ....... =~:: .... ~&quot; :~; The use of prepositions can also affect which quantifier to use. For example, &quot;After all the episodes, the patient received dobutamine&quot; is ambiguous in regards to whether the dobutamine is given once during the surgery, or given after each episode. In contrast, the sentence &quot;In all the episodes, the patient received dobutamine.&quot; does not have this problem.</Paragraph> <Paragraph position="9"> The current system looks at the particular preposition (i.e., &quot;before&quot;, &quot;after&quot;, or &quot;in&quot;) before selecting the appropriate quantifier.</Paragraph> </Section> <Section position="4" start_page="103" end_page="104" type="sub_section"> <SectionTitle> 4.4 Examples of a Single Quantifier </SectionTitle> <Paragraph position="0"> Given the four propositions, &quot;After intubation, Mrs. Doe had tachycardia&quot;, &quot;After skin incision, Mrs. Doe had tachycardia&quot;, &quot;After start of bypass, Mrs. Doe had tachycardia', and &quot;After coming off bypass, Mrs. Doe had tachycardia.&quot;, the algorithm first identifies roles with similar entities, ARG1, PRED, ARG2 and removes them from further quantification processing while the distinct entities in the role MODS-TIME, &quot;after intubation&quot;, &quot;after skin incision&quot;, &quot;after start of bypass&quot;, and &quot;after coming off bypass&quot;, are further processed for universal quantification. The role MODS-TIME is further separated into two smaller roles, one role with the prepositions and the other role with different critical points.</Paragraph> <Paragraph position="1"> Since the prepositions are all the same, universal quantification is only applied to the distinct entities in set X, in this case, the four critical points. Queries to the CLASSIC ontology indicate that the entities in set X, &quot;intubation&quot;, &quot;skin-incision&quot;, &quot;startof-bypass&quot;, and &quot;conaing-off-bypass&quot; match all the possible types of the concept critical-point, satisfying the domain ontology context in Section 4.2.</Paragraph> <Paragraph position="2"> Since set D and set X match exactly, generalization and universal quantification can be used to replace the references to these entities: &quot;After each critical point, Mrs. Doe had tachycardia.&quot; The system currently does not.perfor.m generMization omeJ~tities which failed the univeral quantification test.. In such cases, a sentence with conjunction will be generated, i.e., &quot;After intubation and skin incision, Mrs. Doe had tachycardia.&quot; In addition to every, the system generates both when the number of entities in set X is two. In our application, both is used as a universal quantitier under discourse context: &quot;Alice had q)isodes of bradycardia b@)re inductio1~ and start of bypass, h~ both episodes, she received Cefazolin and Phen!llephrine. &quot; When a universal quantifier is under the govern- null ment of negation, each, all, every and both are inappropriate, and any should be used instead. Given that the patient went on bypass without complications, the system should generate &quot;The patient went on bypass without any problem.&quot; In contrast, &quot;The patient went on bypass without every prob/em.V=-~as-~ a,:differeut.-~meani~g; -,Our, :,system-cur=. rently uses any as a universal quantifier when the universal quantification is under the government of negation, such as &quot;The patient denied any drug allergy.&quot;, or &quot;Her hypertension was controlled without any medication.&quot; Currently, the generation of negation sentences about surgery problems and allergies are handled in the content planner. They are not synthesized from multiple negation sentences: &quot;The patient is not allergic to aspirin. The paitent is not allergic to penicillin...&quot;</Paragraph> </Section> </Section> <Section position="7" start_page="104" end_page="104" type="metho"> <SectionTitle> 5 Generation of Multiple Quantifiers </SectionTitle> <Paragraph position="0"> When there are two distinct roles across the propositions, the algorithm tries to use a universal quantifier for one role and an existential quantifier for another.</Paragraph> <Paragraph position="1"> To generate sentences with 33, both entities being referred to must have no proper names; this triggers the use of existential quantifiers. We intentionally ignore the cases where two universal quantifiers are generated in the same sentence. The likelihood for input specifying sentences with W to a text generation system is slim.</Paragraph> <Paragraph position="2"> When generating multiple quantifiers in the same sentence, we differentiate between cases where there is or isn't a dependency between the two distinct roles. Two roles are independent of each other when one is not a modifier of the other. For example, the roles ARG1 and ARG2 in a proposition are independent. In &quot;Each patient is given a high severity rating&quot;, performing universal quantification on the patients (ARG3) is a separate decision from the existential quantification of the severity ratings (ARG2). Similarly, in &quot;An abnormal lab result was seen in each patient with hypertension after bypass&quot;. the quantification operations on the abnormal lab results and the patients can be performed independently. null .... When there isa dependency 'between theroles being quantified, the quantification process of each role might interact because modifiers restrict the range of the entities being modified. We found that when universal quantification occurs in the MODS role, the quantification of PRED and MODS can be performed independently, just as in the cases withou! dependency. Given the input propositions &quot;Alice has II<I in Alice's left arm. Alice has IV-2 in Alice's right arm. &quot;, the distinct roles are ARG2 &quot;IV-i&quot; and &quot;IV-T', and ARG2-MODS &quot;in Alice's left arm&quot; and &quot;in Alice's right arm&quot;. The ARG2-MODS is universally quantified based on domain knowledge that (r) Roles without dependency, V Role-l,3 Role-2 Each patient is given a high severity rating.</Paragraph> <Paragraph position="3"> (r) Roles without dependency, 3 Role-l, 'v' Role-2 An abnormal lab result was seen in each patient geon's name is likely to be known, and the input is likely to be &quot;Dr. Rose operated on Alice&quot;, &quot;Dr~ Rose operated on Bob&quot;, and &quot;Dr. Rose operated on Chris&quot;. Given these three propositions, the entities in ARG1 and PRED are identical, and only with hypertension after bypass, the distinct entities in ARG2, &quot;Alice&quot;, &quot;Bob&quot; and *Roles with depend~flcy, V PRED,-3 MODS ............. Ghns:,~-~ d.~be:;~qua,atffied.... ~Wilih-,-am:a;ppropriate context, the sentence &quot;Dr. Rose operated on each Every patient with a balloon pump had hypertension. null * Roles with dependency, 3 PRED, V MODS Alice has an IV in each arm,.</Paragraph> <Paragraph position="4"> a patient is a human and a human has a left arm and a right arm. In this example, &quot;an IV in each arm&quot;, the decision to generate universal and existential quantified expressions are independent. But in &quot;Every patient with a balloon pump had hypertension&quot;, the existentially quantified expression &quot;with a balloon pump&quot; is a restrictive modifier of its head. In this case, the set D does not include all the patients, but only the patients &quot;with a balloon pump&quot;. When computing set D for universal quantification, the algorithm takes this extra restriction into account by eliminating all patients without such a restriction.</Paragraph> <Paragraph position="5"> Once a role is universally quantified and the other is existentially quantified, our algorithm replaces both roles with the corresponding quantified expressions.</Paragraph> <Paragraph position="6"> Figure 3 shows the sentences with multiple quantitiers generated by applying our algorithm.</Paragraph> <Section position="1" start_page="104" end_page="104" type="sub_section"> <SectionTitle> 5.1 Ambiguity Revisited </SectionTitle> <Paragraph position="0"> In Section 4.3, we described how to minimize the ambiguity between distributive and collective readings when generating universal quantitiers. What about the scope ambiguity when there are muhiple quantifiers in the same sentence? If we look at the roles which are being universally and existentially quantified in our examples in Figure 3, it is interesting to note that the universal quantifiers always have wider scope than the existential quantifiers. In the first, example, ,the.scope: order is Vpatient~highseverity-rating, the second example is Vpatient31abresult, the third is Vpatient3balloon-pump, and the fourth is Varm3IV. The scope orderings are all V3.</Paragraph> <Paragraph position="1"> \Vhat happens if a sentence contains an existential quantifier which has a wider scope than a universal quantifier? In &quot;A suryeon operated on each patient.&quot;, tile normal reading is Vpatienl3surgeon.</Paragraph> <Paragraph position="2"> 13ut~ if the existentially quantified noun phrase &quot;'a surgeon&quot; refers to tile same surgeon, as in 3surgeonVpatient. tlle system would generate &quot;(A particular/The same) surgeon operated on each patient.&quot; In an applied generation system, the surpatient&quot; will be generated. If the name of the surgeon is not available but the identifiers for the surgeon entities across the propositions are the same, the system will generate &quot;The same surgeon operated on each patient.&quot; As this example indicates, when 3 has a wider scope than V, the first step in our algorithm (described in Section 4.1), identifying roles with distinct entities, would eliminate the roles with identical entities from further quantification processing. Based on our algorithm, the sentences with 3V readings are taken care of by the first step, identifying roles with distinct entities, while V3 cases are handled by quantification operations for multiple roles, as described in Section 5.</Paragraph> <Paragraph position="3"> In Section 4.3, we mentioned that it is important to know exactly what blood products are used in our application. As a result, the system would not generate the sentence &quot;Each patient received a blood product.&quot; when the input propositions are &quot;Alice received packed red blood cells. Bob received platelets.</Paragraph> <Paragraph position="4"> Chris received platelets.&quot; Even though tim conjoined entities can be generalized to &quot;blood product&quot;, this quantification operation would violate our precondition for using existential quantifiers: the descriptions for each of the conjoined entities must be indistinguishable. Here, one is &quot;red blood cells&quot; and tile others are &quot;platelets&quot;. Given these three propositions, the system would generate &quot;Alice received packed red blood cells, and Bob and Chris, platelets.&quot; based on the algorithm described in (Shaw. 1998). If in our domain the input propositions could be &quot;'Alice received blood-product-1. Bob received bloodproduct-2. Chris received blood-product-2.&quot;, where each instance of blood-product-n could be realized as &quot;blood product&quot;, then the system would generate &quot;Each patient received a blood product.&quot; since the description of conj0ined entities are not dist~inguish able at the surface level.</Paragraph> </Section> </Section> <Section position="8" start_page="104" end_page="105" type="metho"> <SectionTitle> 6 Conclusion </SectionTitle> <Paragraph position="0"> We have described the quantification operators that can make the text more concise while preserving the original semantics in the input propositions. Though we would like to incorporate imprecise quantifiers such as few. many, some into our system because they have potential to drastically reduce the text.</Paragraph> <Paragraph position="1"> further, these quantifiers do not, have the desired property ill which the readers can recover the exact.</Paragraph> <Paragraph position="2"> entities in the input propositions. The property of preserving the original semantics is very important since it guarantees that even though the surface expressions are modified, the information is preserved. This property allows the operators to be domain independent and reusable in different natural language Norman Creaney. 1999. Generating quantified logi: cal forms from raw data. In Proe. of the ESSLLI99 Workshop on the Generation of Nominal Expressions. null M. Dalal~ S. Feiner, K. McKeown, D. Jordan, generation systems. B. Allen, and Y. alSafadi. 1996. MAGIC: An We have described: an. algo_r.itlma :which.sy.stemati .............. e:~cpertimeeataL:aystem..for: genetattiag~ .multimedia cally derives quantifiers from input propositions, discourse history and ontological information. We identified three types of information from the discourse and ontology to determine if a universal quantifier can be applied. We also minimnized the ambiguity between distributive and collective readings by selecting an appropriate universal quantifier. Most importantly, for multiple quantifiers in the same sentence, we have shown how our algorithm generates different quantifed expressions for different scope orderings. null</Paragraph> </Section> class="xml-element"></Paper>