File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/83/j83-1001_metho.xml
Size: 32,082 bytes
Last Modified: 2025-10-06 14:11:35
<?xml version="1.0" standalone="yes"?> <Paper uid="J83-1001"> <Title>Paraphrasing Questions Using Given and New Information 1</Title> <Section position="5" start_page="0" end_page="0" type="metho"> <SectionTitle> 4 For example, see Prince 1979 for a discussion of various </SectionTitle> <Paragraph position="0"> usages of &quot;given/new&quot;.</Paragraph> </Section> <Section position="6" start_page="0" end_page="0" type="metho"> <SectionTitle> 2 American Journal of Computational Linguistics, Volume9, Number 1, January-March 1983 </SectionTitle> <Paragraph position="0"> Kathleen R. McKeown Paraphrasing Questions Using Given and New Information refers to information the speaker assumes is present in the data base. Information functioning in the role just described has been termed &quot;given&quot;.</Paragraph> <Paragraph position="1"> &quot;New&quot; labels all information in the sentence that is presented as not retrievable from context. In the declarative, elements functioning in asserting information that the listener is presumed not to know are called new. In the question, elements functioning in conveying what the speaker wants to know (i.e., what she/he doesn't know) represent information the speaker presumes the listener is not already aware of. Firbas 1974 identifies additional functions in the question.</Paragraph> <Paragraph position="2"> Of these, (ii) is used here to augment the interpretation of new information. He says (p. 31): (i) it indicates the want of knowledge on the part of the inquirer and appeals to the informant to satisfy this want.</Paragraph> <Paragraph position="3"> (ii) \[a\] it imparts knowledge to the informant in that it informs him what the inquirer is interested in (what is on her/his mind) and \[b\] from what particular angle the intimated want of knowledge is to be satisfied.</Paragraph> <Paragraph position="4"> Although word order vis-a-vis these and related distinctions has been discussed in light of the declarative sentence, less has been said about the interrogative form. Halliday 1967 and Krizkova s are among the few to have analyzed the question. Despite the fact that they arrive at different conclusions, 6 the two follow similar lines of reasoning. Krizkova argues that both the wh-item of the wh-question and the finite verb (e.g., &quot;do&quot; or &quot;be&quot;) of the yes/no question point to the new information to be disclosed in the response.</Paragraph> <Paragraph position="5"> These elements, she claims, are the only unknowns to the questioner. Halliday, in discussing the yes/no question, also argues that the finite verb is the only unknown. The polarity of the text is in question and the finite element indicates this.</Paragraph> <Paragraph position="6"> In this paper the interpretation of the unknown elements in the question as dfined by Krizkova and Halliday is followed. The wh-items, in defining the questioner's lack of knowledge, act as new information. Firbas's analysis of the functions in questions is used to further elucidate the role of new information in questions. The remaining elements are given information. They represent information assumed by the questioner to be true of the data base domain. This 5 Summary by Firbas 1974 of the untranslated article &quot;The Interrogative Sentence and Some problems of the So-called Functional Sentence Perspective (Contextual Organization of the Sentence),&quot; NASA Rec. 4, 1968.</Paragraph> <Paragraph position="7"> 6 It should be noted that Halliday and Krizkova discuss the unknowns in the question in order to define the theme and rheme of a question. Although they agree about the unknowns for the questioner, they disagree about which elements function as theme and which function as rheme. A full discussion of their analysis and conclusions is given in McKeown 1979.</Paragraph> <Paragraph position="8"> labeling of information within the question will allow the construction of a natural paraphrase, avoiding ambiguity.</Paragraph> </Section> <Section position="7" start_page="0" end_page="0" type="metho"> <SectionTitle> 5. Formulation </SectionTitle> <Paragraph position="0"> Following the analysis described above, the CO-OP paraphraser breaks down questions into given and new information. More specifically, an input question is divided into three parts, of which (2) and (3) form the new information.</Paragraph> <Paragraph position="1"> 1. given information 2. lack of knowledge (ii\[a\] from Firbas above) 3. angle (ii\[b\] from Firbas above) In terms of the question components, part (2) is indicated by the question with no subclauses 7 as it defines the lack of knowledge of the hearer. Part (3) is indicated by the direct and indirect modifiers of the interrogative words as they define the angle from which the question was asked. They identify the attributes of the missing information for the hearer. Part (1) is formed from the remaining clauses. As an example, consider question (D): (D) Which division of the computing facility works on projects using oceanography research? Following the outline above, part (2) of the paraphrase will be the question minus the subclauses: &quot;Which division works on projects?&quot; Part (3), the modifiers of the interrogative words, will be &quot;of the computing facility&quot;, which modifies &quot;which division&quot;.8 The remaining clause &quot;projects using oceanography research&quot; is considered given information. The three parts can then be assembled into a natural sequence: (E) Assuming that there are projects using oceanography research, which division works on those projects? Look for a division of the computing facility. 9 Information belonging to each of the three categories occurred in question (D). If one of these types of information is missing, the question will be presented minus the initial or concluding clauses. Only part (2) of the paraphrase will invariably occur. Note that this means that if there are no clauses in the original question corresponding to parts (1) and (2) (i.e., the question contains no relative clauses, prepositional phrases, 7 Here, subclauses are defined as relative clauses, prepositional phrases, and adjectival phrases.</Paragraph> <Paragraph position="2"> 8 Note that this phrase also identifies a presupposition of the questioner. For the paraphrase, however, its function to precisely specify what the questioner is interested in (which is new information for the hearer) is of greater importance.</Paragraph> <Paragraph position="3"> 9 This example, as well as sample questions and paraphrases that follow, were taken from actual sessions with the paraphraser. Question (A) and its possible paraphrases (B) and (C) were not run on the system.</Paragraph> <Paragraph position="4"> American Journal of Computational Linguistics, Volume 9, Number 1, January-March 1983 3 Kathleen R. McKeown Paraphrasing Questions Using Given and New Information or adjectival phrases), the paraphrase may be the same as the original question.</Paragraph> <Paragraph position="5"> If more than one clause occurs in a particular category, the question will be further splintered. Additional given information is parenthesized following the &quot;assuming that ...&quot; clause. Example (F) below illustrates the paraphrase for a question containing several clauses of given information and no clauses defining specific attributes of the missing information. Clauses containing information characterized by category (3) will be presented as separate sentences following the stripped-down question. (G) below demonstrates a paraphrase containing more than one clause of this type of information.</Paragraph> <Paragraph position="6"> (F) Q: Which users work on projects in oceanography that are sponsored by NASA? P: Assuming that there are projects in oceanography (those projects are sponsored by NASA), which users work on those projects? (G) Q: Which programmers in superdivision 5000 from the ASD group are advised by Thomas Wirth? P: Which programmers are advised by Thomas Wirth? Look for programmers in superdivision 5000. The programmers must be from the ASD group.</Paragraph> </Section> <Section position="8" start_page="0" end_page="0" type="metho"> <SectionTitle> 6. Implementation Overview </SectionTitle> <Paragraph position="0"> The paraphraser's first step in processing is to reform the parse tree it is given so that the main verb occurs as the root of the new tree. This is done to simplify the identification of given and new information in the parse. The tree is then divided into three separate trees reflecting the division of given and new information in the question. The design of the tree allows for a simple set of rules that flatten the tree. The final stage of processing in the paraphraser is translation.</Paragraph> <Paragraph position="1"> In the translation phase, labels in the parser's representation are translated into their corresponding words. During this process, necessary transformations of the grammar are performed upon the string.</Paragraph> <Section position="1" start_page="0" end_page="0" type="sub_section"> <SectionTitle> 6.1 The phrase structure tree </SectionTitle> <Paragraph position="0"> In its initial processing, the paraphraser transforms the parser's representation into one that is more convenient for generation purposes. The resultant structure is a tree that highlights certain syntactic features of the question. This initial processing gives the paraphraser some independence from the CO-OP system.</Paragraph> <Paragraph position="1"> Were the parser's representation changed or the component moved to a new system, only the initial processing phase would need to be modified.</Paragraph> <Paragraph position="2"> The paraphraser's phrase structure tree uses the main verb of the question as the root node of the tree.</Paragraph> <Paragraph position="3"> The subject of the main verb is the root node of the left subtree, the object (if there is one) the root node of the right subtree. In the current system, the use of binary relations in the parser's representation 10 creates the illusion that every verb or preposition has a sub-ject and object. The paraphraser's tree does allow for the representation of other constructions should the incoming language use them.</Paragraph> <Paragraph position="4"> Note that the use of binary relations in the incoming parse tree to represent the verbs and prepositions of a sentence means that modifiers of verbs are represented as modifiers of their objects (and thus hang off the object in the paraphraser's reformed tree). While this is not the usual interpretation of questions using such constructions, it functions adequately for both CO-OP and the paraphraser as illustrated by a hypothetical paraphrase for such a question, shown below in (H): (H) Q: Which programmers worked on oceanography projects in 1972? P: Assuming that there were oceanography projects in 1972, which programmers worked on those projects? Each of the paraphrase subtrees represents other clauses in the question. Both the subject and the object of the main verb will have a subtree for each other clause it participates in. If a noun in one of these clauses also participates in another clause in the sentence, it too will have subtrees.</Paragraph> <Paragraph position="5"> As an example, consider the question: &quot;Which active users advised by Thomas Wirth work on projects in area 3?&quot; The phrase structure tree used in the paraphraser is shown in Figure 1. Since &quot;work on&quot; is identified as the main verb of the question by the parser, it will be the root node of the tree. &quot;users&quot; is root of the left subtree, &quot;projects&quot; of the right. Each noun participates in one other clause and therefore has one subtree. Modifiers are closely bound to the noun they modify and are treated as properties of the noun (i.e., each node in the tree that is modified has a prop-erty called &quot;modifiers&quot; whose value is any adjectival or noun modifier). In Figure 1, modifiers are shown as part of the node label for clarity. Subtree nodes (the leaves of Figure 1) have three pieces of information associated with them: * the relation between the node and its parent, * the noun phrase the node represents, and * an indication of whether the node functions as subject or object in the clause.</Paragraph> <Paragraph position="6"> 10 See Kaplan 1979 for a description of Meta Query Language, or MQL.</Paragraph> <Paragraph position="7"> 4 American Journal of Computational Linguistics, Volume 9, Number 1, January-March 1983 Kathleen R. McKeown Paraphrasing auestions Using Given and New Information</Paragraph> </Section> <Section position="2" start_page="0" end_page="0" type="sub_section"> <SectionTitle> 6.2 Dividing the tree </SectionTitle> <Paragraph position="0"> The constructed tree is computationally suited for the three-part paraphrase. The tree is flattened after it has been divided into subtrees containing given information and the two types of new information. The splitting of the tree is accomplished by first extracting the topmost smallest portion of the tree containing the wh-item. At the very least, this will include the root node plus the left and right subtree root nodes. This portion of the tree is the stripped-down question. The clauses that define the particular aspect from which the question is asked are found by searching the left and right subtrees for the wh-item or questioned noun.</Paragraph> <Paragraph position="1"> The subtree whose root node is the wh-item contains these clauses. Note that this may be the entire left or right subtree or may only be a subtree of one of these.</Paragraph> <Paragraph position="2"> The remainder of the tree represents given information. Figure 2 illustrates this division for the previous example.</Paragraph> <Paragraph position="3"> with A the left subtree and B the right, then the following rules define the flattening process:</Paragraph> <Paragraph position="5"> In other words, the top level of the tree (shown on the left in Figure 3) is linearized by an in-order traversal while each of its subtrees (shown on the right in Figure 3) is linearized by a pre-order traversal. In the example shown in Figure 2, part (2) of the tree corresponds to the top level of the tree and will undergo in-order linearization, and parts (1) and (3) are the subtrees, which will be linearized by a pre-order traversal. The use of two traversals to linearize the tree stems from the fact that different types of information are stored at nodes at different levels in the tree. As a node in a subtree has three pieces of information associated with it, one more rule is required to expand a node. A node consists of:</Paragraph> <Paragraph position="7"> where arc-label is the label of a binary relation in the input parse tree (i.e., a verb or preposition) and set-label is the label of a set in the input parse (i.e., noun phrase). The input parse is in MQL representation, which consists of sets and binary relations between them. Subject/object indicates whether the sub-node noun phrase functions as subject or object in the clause; it is used by the subject-aux transformation and does not apply to the expansion rule. In Figure 2, the leaves of the tree carry these three pieces of information. For example, the leftmost leave has arc-label advised by, set-label Thomas Wirth, and is labeled as American Journal of Computational Linguistics, Volume 9, Number 1, January-March 1983 5 Kathleen R. McKeown Paraphrasing Questions Using Given and New Information the object of the relation. The following rule expands a subtree node:</Paragraph> </Section> </Section> <Section position="9" start_page="0" end_page="4" type="metho"> <SectionTitle> NODE -~ ARC-LABEL SET-LABEL </SectionTitle> <Paragraph position="0"> The tree of given information is flattened first. It is part of the left or right subtree of the phrase structure tree and therefore is flattened by a pre-order traversal. It is during the flattening stage that the words &quot;Assuming that there \[be\] ...&quot; are inserted to introduce the clause of given information. &quot;be&quot; will agree with the subject of the clause. Following these rules, the tree of given information in Figure 2 would be flattened by a pre-order traversal yielding &quot;projects in area #6&quot; (R' A t arc-label set-label). After the &quot;Assuming that&quot; clause is inserted, this portion of the paraphrase is &quot;Assuming that there be projects in area #6&quot;. If there is more than one clause, parentheses are inserted around the additional ones.</Paragraph> <Paragraph position="1"> The tree representing the stripped-down question is flattened next, using the in-order traversal. Applying this process to Part (2) of the tree in Figure 2 yields the phrase &quot;wh active users work on projects&quot; (A R B). (In final processing stages, the correct demonstrative (&quot;those&quot; or &quot;that&quot;) is selected to modify nouns already mentioned in the first part of the paraphrase.) The tree that represents modifiers of the questions noun is linearized to follow these phrases. A pre-order traversal of this portion of the tree in Figure 2 yields &quot;users advised by Thomas Wirth&quot; (R t A T arc-label set-label). Any modifiers of a noun (here, &quot;active&quot;) are omitted in this part of the paraphrase if they have already been mentioned. The phrase &quot;Look for&quot; is inserted before the first clause of modifiers.</Paragraph> <Paragraph position="2"> Two transformations are applied during the flattening process. They are wh-fronting and subject-aux inversion. Other transformations are applied following the flattening process to produce sentences in final grammatical form.</Paragraph> <Section position="1" start_page="0" end_page="4" type="sub_section"> <SectionTitle> 6.4 Transformations </SectionTitle> <Paragraph position="0"> The grammar used in the paraphrase is a transformational one. In addition to the basic flattening rules described above, the following transformations are The curved lines indicate the ordering restrictions. There are two connected groups of transformations. If wh-fronting applies, then so will do-support, subject-aux inversion, and tense-placement. The second group Input to rule: of transformations is invoked through the application of negation. It includes do-support, contraction, and tense-placement. Has-deletion is not affected by the absence or presence of other transformations. A description of the transformation rules follows. The rules used here are based on analyses described by Akmajian and Heny (1975) and by Cullicover (1976). The rule for wh-fronting is specified as follows, where SD stands for structural description and SC, structural changes. Each rule is followed by an example input string and the string after it has undergone the transformation. The full tree for the string is not shown, but the string is labeled by markers in the SD.</Paragraph> <Paragraph position="2"> The first step in the implementation of wh-fronting is a search of the tree for the wh-item. A slightly different approach is used for paraphrasing than would be used if simply generating a question from the input parse. The difference occurs because in the original question the NP to be fronted may be the head noun of some relative clauses or prepositional phrases. If generating, these clauses would be fronted along with the head noun. Since the clauses of the original question are broken down for the paraphrase, it will never be the case when paraphrasing that the NP to be fronted also dominates relative clauses or prepositional phrases. For this reason, the applicability of wh-fronting is testing for and is applied in the flattening process of the stripped-down question. Note that the phrase markers (or categories) of each word are retained as the tree is flattened and thus the SD's can be matched against both the tree and its linearized version. If wh-fronting applies, only one word need be moved to the initial position.</Paragraph> <Paragraph position="3"> The paraphraser is capable of generating English from the input as well as paraphrasing (see Section 7).</Paragraph> <Paragraph position="4"> When generation is being done, the applicability of wh-fronting is tested for immediately before flattening.</Paragraph> <Paragraph position="5"> If the transformation applies, the tree is split. The subtree of which the wh-item is the root is flattened separately from the remainder of the tree and is attached in fronted position to the string resulting from flattening the other part.</Paragraph> <Paragraph position="6"> After wh-fronting has been applied, do-support is invoked. In CO-OP, the underlying representation of I I past plur work on? Subject-aux inversion is activated immediately afterwards. Again, if wh-fronting is applied, subject-aux inversion will apply also. The rule is: past plur work on? Tense-placement follows subject-aux inversion* Tense, number, and negation (if present) are attributes of all verbs in the parser's representation* When an auxiliary is generated, the tense, number, and negation are moved from the verb to the auxiliary* Formally:</Paragraph> <Paragraph position="8"> Some transformational analyses propose that wh-fronting and subject-aux inversion apply to the relative clause as well as the question. In the CO-OP paraphraser, the head-noun is properly positioned by the flattening process and wh-fronting need not be used.</Paragraph> <Paragraph position="9"> Subject-aux inversion, however, may be applicable. In cases where the head noun of the clause is not its subject, subject-aux inversion results in the proper order* The rule for negation is tested during the translation phase of execution. It has been formalized as: 'wh students' 'pres plur hav~ ' no ' fadvisors?l In the CO-OP representation, an indication of negation is carried on the object of a binary relation (see Kaplan 1979)* When generating an English representation of the question, it is possible in some cases to express negation as modification of the noun (see question (H) below)* In all cases, however, negation can be indicated as part of the verb (see version (I) of question (H)). Therefore, when the object is marked as negative, the paraphraser moves the negation to become part of the verbal element* (H) Which students have no advisors? (I) Which students don't have advisors? In English, the negative marker is attached to the auxiliary of the verbal element and, therefore, as was the case for questions, an auxiliary must be generated* Do-support is used. The rule for do-support after American Journal of Computational Linguistics, Volume 9, Number 1, January-March 1983 7 Kathleen R. McKeown Paraphrasing Questions Using Given and New Information negation differs from the one used after wh-fronting.</Paragraph> <Paragraph position="10"> They are presented this way for clarity, but could have been combined into one rule.</Paragraph> <Paragraph position="11"> Tense-placement, as described above, moves the tense, number, and negation from the verb to the auxiliary verb. The cycle of transformations invoked through application of negation is completed with the contraction transformation. The statement of the contraction transformation is:</Paragraph> <Paragraph position="13"> where # indicates that the result must be treated as a unit for further transformations. The morphology routines will combine the result to produce &quot;don't&quot;. corrective response that could be generated by the paraphraser if (J) were asked: (J) Which programmers in division 3 work on projects in oceanography? (K) I don't know of any projects in oceanography.</Paragraph> <Paragraph position="14"> Alternative suggestions are also used by the CO-OP system when the direct response to the user's question is negative. If an incorrect presupposition is removed from a question, the resulting question may no longer have a negative response. 11 In such cases, CO-OP suggests the wider class question to the user as a possible interest. CO-OP passes the MQL representing this question to the paraphraser, which generates the English for the suggestion. A sequence like (J), (K) above might be followed by the alternative suggestion (L): (L) But you might be interested in programmers in division 3 that work on any projects.</Paragraph> <Paragraph position="15"> For both types of responses, the paraphraser generates the response using the paraphrase functions with minor differences. The flattening process for generation differs from that used for paraphrases in that the tree is not divided into subtrees representing given and new information and, therefore, the tree is flattened as a whole. The transformational grammar also applies to the generation process, with the one difference being the point at which the applicability of wh-fronting is tested for (described in Section 6.4). Other than these changes and the use of different leading phrases (e.g., &quot;But you might be interested in ...&quot;), the generation process is the same as the paraphraser process. The generation function is general enough that it could be used for other types of responses in cases when something other than a direct response is needed.</Paragraph> </Section> </Section> <Section position="10" start_page="4" end_page="4" type="metho"> <SectionTitle> 8. Related Research 7. Other Features of the Paraphraser </SectionTitle> <Paragraph position="0"> The paraphraser is used for a second purpose in addition to paraphrasing. It can generate an English version of the parser's representation as well as paraphrase in the three-part form. This function uses the same procedures and grammar as the three-part paraphraser, but the tree is not split into three separate trees before being flattened.</Paragraph> <Paragraph position="1"> In CO-OP, generation is used to produce alternative suggestions and corrective responses. A corrective response is used to correct the user's false presuppositions. When an existential presupposition encoded in the question is incorrect, the portion of MQL representing the failed presupposition (this is determined by CO-OP) is passed to the paraphraser, which generates the corrective response. For example, (K) below is a At the time of the CO-OP paraphraser implementation, two main other paraphrasers had been developed and implemented for data base question-answering systems: null * PLANES, Waltz et al. 1978; * RENDEZVOUS Version 1, Codd 1978.</Paragraph> <Paragraph position="2"> Both systems used templates to form the paraphrases.</Paragraph> <Paragraph position="3"> Templates are canned English phrases (or sentences) containing slots that may be filled with different words to produce a variety of full English phrases.</Paragraph> <Paragraph position="4"> The PLANES system generates the paraphrase from the formal data base query using templates. The process involves three specific actions. English words are substituted for any abbreviations or code names in the ll See Kaplan 1979 for details on determining the most appropriate alternative suggestion.</Paragraph> </Section> <Section position="11" start_page="4" end_page="4" type="metho"> <SectionTitle> 8 American Journal of Computational Linguistics, Volume 9, Number 1, January-March 1983 </SectionTitle> <Paragraph position="0"> Kathleen R. McKeown Paraphrasing Questions Using Given and New Information data base query, using a table look-up. A single appropriate paraphrase template is selected for use based on the query, and the slots in the template are then filled with words and phrases from the query. The major effort in designing this kind of system is in the formation, by hand, of templates suitable for the particular data base and for the types of questions that can be asked. An example of an English question and the PLANES paraphrase for it are shown below in (M): (M) Q: How many flights did plane 3 make in Jan 73? P: PLANES searches the MONTHLY FLIGHT and MAINTENANCE SUMMARIES and returns: The value of TOTAL FLIGHTS for plane SERIAL #3 during January 1973.</Paragraph> <Paragraph position="1"> The RENDEZVOUS system also generates the paraphrase from the formal query using templates, although it is slightly more sophisticated than Waltz's. There are three parts to generation, and two types of templates are used. A header template corresponding to the type of query is chosen first. There are three types of queries in the system (FIND, EXIST, COUNT), of which FIND occurs most frequently. The header for FIND is PRINT THE ... EVERY .... where the dots must be filled in. The second part to the paraphrase is the target list. It specifies the attributes requested by the user and is supplied by doing a table look-up on the attribute. The third part of the paraphrase is called the body. It is formed by extracting templates from tables, associated with particular items in the query, that specify restrictions on the requested values. An example of a query and the paraphrase generated by RENDEZVOUS is shown in (N) below.</Paragraph> <Paragraph position="2"> (N) Q: I want to find certain projects. Pipes were sent to them in Feb. 1975.</Paragraph> <Paragraph position="3"> P: Print the name of every project to which a shipment of a part named pipe was sent during February 1975.</Paragraph> <Paragraph position="4"> The goals of the RENDEZVOUS generation component are important ones. The generated English must be unambiguous, easy to understand, discriminating, and not misleading (Codd 1978). Instead of developing a general solution to achieve these goals, however, the research seems to be concentrated on particular examples which don't meet these criteria. This results in part from the use of templates. The templates must be constructed beforehand for a particular data base, and great care must be taken to choose phrases that can be easily patched together with a variety of other phrases. Unforeseen interaction between juxtaposed phrases is a problem that frequently arises. Such an approach necessitates looking at particular examples, instead of the general framework.</Paragraph> <Paragraph position="5"> In both of these systems, the use of templates means that the major effort in developing the system must be done by hand in formatting the English phrases. All questions that will be asked must be anticipated ahead of time, and although the systems can be extended by adding new templates, undesirable interactions between new and old templates must be specifically avoided, and each new required addition does not ease the addition of subsequent templates. Note that this means coverage in a template system is also difficult to specify.</Paragraph> <Paragraph position="6"> The use of a grammar in the CO-OP paraphraser makes it more flexible than these earlier paraphrasers: * less work must be done by hand in formulating the system, * interactions between templates are not a problem since the grammar determines how to combine words and phrases in an acceptable way, and * the system is capable of handling new questions for which it has not been explicitly prepared, as long as they fall within the syntactic range of the system.</Paragraph> <Paragraph position="7"> The paraphraser's ability to perform the generation task described in the previous section nicely illustrates its flexibility. Note furthermore that the CO-OP paraphraser specifically addresses the problems of disambiguating relative clause modification in a general way and of generating a paraphrase that differs from the original question on a theoretical basis, issues not addressed by either the PLANES or the RENDEZVOUS paraphraser.</Paragraph> </Section> class="xml-element"></Paper>