File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/97/w97-1305_metho.xml
Size: 29,487 bytes
Last Modified: 2025-10-06 14:14:48
<?xml version="1.0" standalone="yes"?> <Paper uid="W97-1305"> <Title>Resolving Anaphoric References on Deficient Syntactic Descriptions</Title> <Section position="3" start_page="30" end_page="30" type="metho"> <SectionTitle> 2 Notions of Robustness </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="30" end_page="30" type="sub_section"> <SectionTitle> 2.1 Robustness in Natural Language Processing </SectionTitle> <Paragraph position="0"> In natural language processing in general, the robustness issue comprises the ability of a software system to cope with input that gives rise to deficient descriptions at some descriptional layerJ More or less implicit is the assumption that the system exhibits some kind of monotonic behaviour: the less deficient the description, the higher the quality of the output (Menzel, 1995).</Paragraph> <Paragraph position="1"> Following Menzel further, this intuitive characterization may be refined. Processing should exhibit autonomy in the sense that complete failures at one stage of analysis should not cause complete failures at other stages of analysis or even a failure of the overall processing. Moreover, the processing model should ideally employ some kind of interaction between different stages of analysis: deficiency at one stage of analysis should be compensated by the information gained at other stages.</Paragraph> </Section> <Section position="2" start_page="30" end_page="30" type="sub_section"> <SectionTitle> 2.2 Robustness and Anaphor Resolution </SectionTitle> <Paragraph position="0"> In the light of the above description, the robustness requirement for the anaphor resolution task may be rendered more precisely. In the aforesaid operational approaches, a sequential processing model is followed according to which anaphor resolution is performed by referring to the result of an already completed syntactic analysis. This architecture, however, tacitly ignores evidence for structural disambiguation that may be contributed by strong expectations at the referential layer (Stuckardt, 1996a). In terms of the general goals of robust processing, this means that, since there is no interaction, the robustness requirement merely shows up in form of the monotonicity and autonomy demands: the anaphor resolution module has to cope with deficient or shallow syntactic information. Besides the trivial way to achieve this kind of robustness by simply not exploiting deficient syntactic descriptions, the following two models may be followed: * the shallow description model: by exploiting heuristic rules to reconstruct syntactic description, the anaphor resolution strategies are 1The deficiency may result either because the input itself is deficient, or due to shortcomings of the processing resources, e.g. lexicon, grammar/parser, or semantic/pragmatic disambiguation.</Paragraph> <Paragraph position="1"> adapted to shallow input data which are never defective. 2 * the deficient description model: by extending anaphor resolution strategies to work on a possibly ambiguous or incomplete description, syntactic evidence is exploited as far as available.</Paragraph> <Paragraph position="2"> In contrast to the approach of Kennedy and Boguraev, which is based on the shallow description model, the subsequent sections develop two methods that follow the deficient description model. At first, a new partly heuristic approach will be described.</Paragraph> <Paragraph position="3"> Secondly, a non-heuristic algorithm will be specified which establishes the conceptually superior degree of robustness through interaction: it makes available the results of anaphor resolution for syntactic disambiguation.</Paragraph> </Section> </Section> <Section position="4" start_page="30" end_page="31" type="metho"> <SectionTitle> 3 Fragmentary Syntax </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="30" end_page="30" type="sub_section"> <SectionTitle> 3.1 Phenomena </SectionTitle> <Paragraph position="0"> The main phenomena which give rise to structural ambiguity of syntactic descriptions are uncertainty of syntactic function (involving subject and direct object) and attachment ambiguities of prepositional phrases, relative clauses, and adverbial clauses. In the example Peter observes the owner of the telescope with it. depending on the availability of disambiguating information, it may be uncertain whether the underlined prepositional phrase with it should be interpreted adverbially or attributively. From the configurational perspective, these ambiguities give rise to fragmentary syntactic descriptions which consist of several tree-shaped connected components. With the exception of the topmost tree fragment, all components correspond to a syntagma of type PP, S, or NP whose attachment or role assignment failed.</Paragraph> <Paragraph position="1"> In addition, cases in which no reading exists give rise to fragmentary syntactic descriptions comprising the constituents whose combination failed due to constraint violation.</Paragraph> </Section> <Section position="2" start_page="30" end_page="31" type="sub_section"> <SectionTitle> 3.2 Fragmentary Syntax and Anaphor Resolution </SectionTitle> <Paragraph position="0"> Among the anaphor resolution strategies potentially affected by fragmentary syntax are heuristics as well as constraints. Preference criteria like salience factors and syntactic parallelism are not affected by 2Here, the monotonicity demand of intuitive robustness virtually vanishes, since there is no longer a syntactic input prone to deficiency.</Paragraph> <Paragraph position="1"> all types of syntactic defects. Moreover, there is a plethora of heuristics which do not rely on syntactic function or structure. Structural coindexing constraints, however, may lose evidence in all above cases of fragmentary syntax. Since they are known to be of central importance to the antecedent filtering phase of operational anaphor resolution approaches, the subsequent discussion focuses on the impact of deficient surface structure description to this class of restrictions.</Paragraph> <Paragraph position="2"> According to the Government and Binding Theory of Chomsky, the core of the syntactic coindexing restrictions is stated as follows Definition 1 (binding principles) (A ) A reflexive or reciprocal is bound in its binding category.</Paragraph> <Paragraph position="3"> (B) A pronominal is free (i.e. not bound) in its binding category.</Paragraph> <Paragraph position="4"> (C) A referring expression is free in any domain. where binding category denotes the next dominator containing some kind of subject (Chomsky, 1981), and binding is defined as coindexed and ccommanding: null Definition 2 (thec-command relation) Surface structure node X c-commands node Y if and only if the next &quot;branching node&quot; which dominates X also dominates Y and neither X dominates Y, Y dominates X nor X = Y.</Paragraph> <Paragraph position="5"> A further structural well-formedness condition, commonly named i-within-i filter, rules out &quot;referentially circular&quot; coindexings, i.e. configurations matching the pattern \[c~ ... Ill ... \]i\]i.</Paragraph> <Paragraph position="6"> In the above example, the latter restriction comes to an application, licensing a coindexing of telescope and it only if the PP containing it is not interpreted as an attribute to telescope - otherwise, in contradiction to the i-within-i condition, the pronoun would be contained in the NP of the tentative antecedent. Hence, if the PP attachment ambiguity has not been resolved prior to anaphor resolution, the fragmentary syntactic description does not contribute the configurational evidence which is necessary for definitely confirming antecedent candidate telescope.</Paragraph> </Section> </Section> <Section position="5" start_page="31" end_page="33" type="metho"> <SectionTitle> 4 Checking Binding Constraints on </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="31" end_page="31" type="sub_section"> <SectionTitle> Fragmentary Syntax 4.1 Basic Observations </SectionTitle> <Paragraph position="0"> The first step towards the verification of binding constraints on fragmentary syntax is suggested by the following observation: If the anaphor as well as the antecedent candidate are contained in the same connected component of the fragmentary syntactic description, no (direct) binding theoretic evidence is lost.</Paragraph> <Paragraph position="1"> In this case, the verification of the binding restrictions of anaphor and antecedent will be possible in a non-heuristic manner, since the necessary positive (---~ binding principle A) and negative (-+ binding principles B, C) syntactic-configurational evidence is entirely available. 3 If, however, the two occurrences belong to different fragments, relevant information may be lost.</Paragraph> <Paragraph position="2"> These considerations give rise to a first solution: to be able to detect which one of the two cases holds, the descriptions of the discourse referent occurrences are supplemented with an attribute which uniquely identifies of the syntactic fragment to which the corresponding NP 4 belongs (e.g. a pointer to the root of the fragment, or a natural number). For a given pair of anaphor a and antecedent candidate 7, the following procedure is applied: If anaphor c~ and candidate ~ occur in the same fragment, verify binding restrictions as in case of unique syntactic description.</Paragraph> <Paragraph position="3"> If they occur in different fragments, consider ,~ a configurationally acceptable candidate for a, but reduce the plausibility score associated with the pair (o~,~/).</Paragraph> <Paragraph position="4"> Consequently, in certain, recognizable cases, the robust binding constraint verification merely yields a heuristic approval of coindexing. The strategy loses a part of its former strictness because configurational evidence is only partially available.</Paragraph> </Section> <Section position="2" start_page="31" end_page="32" type="sub_section"> <SectionTitle> 4.2 Rule Patterns </SectionTitle> <Paragraph position="0"> Even in the disadvantageous case, a closer look at the tree fragments of anaphor and antecedent candidate may reveal additional information. Figure 1 shows rule patterns which exploit this evidence, s</Paragraph> <Paragraph position="2"> In fragmentations matching pattern IF1\], both fragments are constituents which contain the binding categories be(a) and be('7) of the respective occurrences a and '7 of type B. In particular, this implies that the fragments may not be attached in a way that one occurrence locally c-commands the other. Therefore, in both cases, binding principle B is respected, and the coindexing of a and '7 is nonheuristically approved.</Paragraph> <Paragraph position="3"> Conversely, if pattern \[F2\] is matched, &quot;7 is definitively ruled out as the local antecedent prescribed for type A anaphor a: it is impossible to connect the two fragments in a way that &quot;7 locally c-commands a. Here, the fragment of the candidate is only required to contain the branching node of &quot;7, because this suffices to preclude that &quot;7 c-commands a if Fi is embedded in Fj.</Paragraph> <Paragraph position="4"> For certain successive fragment pairs, the parsing result comprises additional information about immediate or transitive embedding. Based on this evidence, further non-heuristic rules (\[FEi\],\[FE2\], \[FE3\], and \[FE4\]) become applicable (Fd = dominating fragment, FC/ = embedded fragment).</Paragraph> <Paragraph position="5"> This list may be supplemented by rules which are based on more subtle configurational case distinctions, and, moreover, by heuristic rules which employ standardized assumptions about typical decision patterns of structural disambiguation. With each of the latter rules, an individual decision plausibility weight may be associated.</Paragraph> <Paragraph position="6"> In an application context, the extension of heuristic rules should be limited to configurations which are known to be of practical relevance. Based on a suitable corpus annotated with syntactic and referential information, it should be possible to determine probthe branching node dominating X according to the c-command definition; the subscript of Xtypev denotes that the binding theoretic class of the occurrence contributed by X is Y E {A, B, C}, e.g. P, yp~ s is a pronominal. x//* indicates that a rule pattern admits/forbids coindexing.</Paragraph> <Paragraph position="7"> abilities for different coindexing configurations by a statistical distribution analysis. These probabilities may then be used to derive promising plausibility weighted rules.</Paragraph> </Section> <Section position="3" start_page="32" end_page="33" type="sub_section"> <SectionTitle> 4.3 An example </SectionTitle> <Paragraph position="0"> The following example illustrates the application of some of the above rules: 6 Der Mann hat den Pr~sidenten besucht, der ihn yon sich iiberzeugte.</Paragraph> <Paragraph position="1"> &quot;The man has the president visited, who him from himself convinced.&quot; Because of the intervening past participle, the relative clause may be interpreted as an attribute to either Mann or Prdsidenten. Hence, syntactic ambiguity arises, yielding a surface structure description which consists of the following two fragments In addition, it is known that the second fragment is embedded in the first. There are three pronominal anaphors to be resolved: the reflexive pronoun sich of type A, the nonreflexive pronoun ihn of type B, and the relative pronoun der of type B.</Paragraph> <Paragraph position="2"> For the reflexive pronoun sieh, the syntactic restrictions may be applied nonheuristically. Candidates der and ihn are contained in the same surface structure fragment. Consequently, binding theoretic evidence is completely available. Since the candidates locally c-command sich, they are both determined to be possible antecedents. The two candidates Mann and Prdsident, however, occur in the other fragment. Hence, it is attempted to apply one of the above rule patterns. Since the reflexive pronoun is of binding theoretic type A, and the fragment in which it occurs contains its binding category (the S node of the relative clause), the Fe fragment pattern of rule \[FE2\] is matched; analogously, the (dominating) fragment containing the type C candidates matches the Fd fragment pattern. Hence, rule \[FE2\] applies: it nonheuristically rules out Mann as well as Priisident. Similarly, for the pronouns ihn and der, type C candidates Mann and Pr~isident are definitively confirmed. Since these anaphors are of type B, the Fe fragment pattern of rule \[FEll is matched. Moreover, the (dominating) antecedent candidate fragment matches pattern Fd. Consequently, \[FE1\] applies and predicts the admissibility of the candidates.</Paragraph> </Section> <Section position="4" start_page="33" end_page="33" type="sub_section"> <SectionTitle> 4.4 An Implementation </SectionTitle> <Paragraph position="0"> The above technique for achieving robustness according to the deficient description model has been integrated into an anaphor resolution system for German text (Stuckardt, 1996b). 7 At present, only nonheuristic rules are employed.</Paragraph> <Paragraph position="1"> In a quantitative evaluation on a corpus of architect biographies (Lampugnani, 1983), the algorithm correctly resolved about 82 per cent of type B pronouns (including possessives). In an idealized test scenario in which correct syntactic readings were manually provided, a precision of 90 per cent was obtained.</Paragraph> <Paragraph position="2"> Hence, on fragmentary syntax, the result quality only decreases by 8 points of percentage. Compared with the 75 per cent achieved by the shallow description approach of Kennedy and Boguraev, this indicates that approaches to robust anaphor resolution which follow the deficient description model may achieve a higher precision. A principled, instructive comparison based on a broader set of text genres has to confirm this improvement.</Paragraph> </Section> </Section> <Section position="6" start_page="33" end_page="36" type="metho"> <SectionTitle> 5 A Complete Algorithm </SectionTitle> <Paragraph position="0"> By the nonheuristic rules of the above method, only those parts of surface structure description are exploited which are valid independently of further disambiguation. It is, however, possible to follow a more principled approach which utilizes configurational evidence that is confined to certain readings. As it will be shown in the following, this may be achieved by tracing the reading dependency of antecedent decisions relying on particular configurations. Through this technique, the results of referential disambiguation can be utilized as evidence for further narrowing structural ambiguity.</Paragraph> <Paragraph position="1"> Tin its principal layout, the algorithm coincides with the one that will be described in section 5.3.</Paragraph> <Section position="1" start_page="33" end_page="34" type="sub_section"> <SectionTitle> 5.1 Dominance Relations in Packed Shared Forests </SectionTitle> <Paragraph position="0"> Following a standardized framework, ambiguous parsing results are henceforth assumed to be represented as packed shared forests (PSFs) (Tomita, 1985). In this representation, structural ambiguity is encoded by packing different derivation variants of input substrings into single interior nodes. Moreover, subtrees common to different readings are allowed to be shared. Formally, such a parse forest can be described as a directed acyclic graph (DAG) with a distinguished topmost element, and leaves corresponding to input words.</Paragraph> <Paragraph position="1"> Given a PSF T with nodes V = {vl,...,Vk}, let P(vl),...,P(Vk) be the respective derivation variants according to packing. Hence,</Paragraph> <Paragraph position="3"> denotes the maximum number of readings (parse trees) represented by T. Consequently, sets of readings may be specified by bit vectors of length n.</Paragraph> <Paragraph position="4"> The application of binding principles crucially rests on the availability of information about dominance relations between parse tree nodes. In ambiguous structure, configurational relations may be confined to certain subsets of readings. The idea now is to qualify dominance information by bit vectors of length n. For each pair (vi,vo) consisting of an interior node vi and an occurrence contributing node vo (usually preterminal), vectors ad(vi, vo) and l~l(vi, Vo) are introduced. Vectors ad(vi, Vo) and ld(vi,Vo) characterize the readings in which vi arbitrarily dominates, or (in the sense of binding theory) locally dominates Vo. Based on these vectors, it will be possible to apply the binding restrictions in a reading sensitive way.</Paragraph> <Paragraph position="5"> By generalizing a technique for unambiguous syntax (Correa, 1988), the reading-qualified dominance information may be precomputed as follows, s The tree traversal process starts at the preterminal nodes which are assumed to be shared among all readings.</Paragraph> <Paragraph position="6"> (If this condition is not satisfied, a preprocessing is performed which, by topdown propagation, determines reading characterization vectors which are then taken for a qualified initialization of the vectors l~/and ad.) Each preterminal node Vp is assigned the following vectors: { (1,...,1), vp ; Vo Jd(v,,,Vo) = l?(vp,Vo) := (0, ,0), vp SAlternatively, this information may be determined on demand of the anaphor resolution task.</Paragraph> <Paragraph position="8"> The computation proceeds bottom-up as follows.</Paragraph> <Paragraph position="9"> Let vi be an interior node for which all descendants of all derivation variants in P(vi) have already been processed. By taking into consideration whether the node delimits a local domain of binding, the vectors are assigned as shown in figure 2 (the operators A and V denote bitwise conjunction and disjunction of vectors, respectively).</Paragraph> <Paragraph position="10"> The computation of the vectors t~d(vi,vo), which characterize the readings in which vi arbitrarily dominates vo, denotes the basic case. The outermost vector disjunction sums over the derivation variants P(vi) which, due to packing, exist for the interior node vi. Bit vector lY(P(vi), m) acts as a filter characterizing the subset of readings in which the m th derivation variant of vi, 1 < m < IP(vi)l, is valid? For each derivation variant, there exists a set D(P(vi), m) of descendants Vd which correspond to nonterminals on the right-hand side of the respective rewriting rule. The nodes that are dominated by these nonterminals are transitively dominated by vi. Hence, the overall result is obtained by recursively summing up the individual contributions of the descendants and qualifying them by conjoining them with p'~P(vi), m).</Paragraph> <Paragraph position="11"> Vectors l~l(vi, vo), which characterize the readings in which vi locally dominates Vo, are computed similarly. The only difference arises if vi delimits a local domain of binding. In this case, the dominance relation computation starts from scratch, i.e. with the zero vector.</Paragraph> </Section> <Section position="2" start_page="34" end_page="35" type="sub_section"> <SectionTitle> 5.2 Binding Principle Verification on Packed Shared Forests </SectionTitle> <Paragraph position="0"> The vectors ad(vi,vo) and ld(vi,Vo) are used to perform referential disambiguation which is sensitive to structural ambiguity. For this purpose, during anaphor resolution, each pair of anaphor a and antecedent candidate '7 (identified with the corresponding preterminal nodes) is assigned a vector g(a,,7) characterizing the readings under which a 9Upon fixation of a particular encoding scheme for reading characterization, vectors lY(P(vi),m) may be computed according to a simple formula.</Paragraph> <Paragraph position="1"> coindexing of a and '7 is configurationally admissible. By taking into account the respective binding principles, F(a, '7) may be determined as follows:</Paragraph> <Paragraph position="3"> where bn(x) represents the set of dominators of a node x which are branching nodes for x in the sense of the c-command definition. The first conjunct specifies the bitvector characterizing the subset of readings under which the binding principle of the anaphor a is (constructively) satisfied. Analogously, the second conjunct describes the (unconstructive) binding principle verification for the antecedent candidate 7. In both cases, branching nodes v have to be considered because they determine the starting point for the application of the dominance information. Since the property of being a branching node relatively to another node may in general depend on packing variants, reading dependency arises. Again, this subtlety is modeled by adding up a set of disjuncts which are qualified by vectors p&quot;~P(v),m), v E bn(x). Here, these vectors characterize the subset of readings in which the property of being a branching node relatively to node x holds for v.</Paragraph> <Paragraph position="4"> The strong (constructive) and weak (unconstructive) verification of binding princi~es is accompl~hed by a conjunction with vectors bps(v, a) and bpw(v,'7), respectively, which, depending on the applicable binding principle, exploit the reading-qualified domination information: b s(vl,v ) := { The sole difference between l~l(vl, v2), if bttype(v2) = A /d(Vl, v2), if bttype(v2) -- B a~d(vl, v2), if bttype(v2) = C (1,..., 1), if bttype(v2) -- A l~l(Vl, v2), if bttype(v2) = B ad(Vl, v2), if bttype(v2) = C the strong and the week 1. For each anaphoric NP a, determine the set of admissible antecedents 5': (a) Verify morphosyntactic or lexical agreement with &quot;7 (b) If the antecedent candidate &quot;7 is intrasentential: by checking that F'(a, 5') C/ (0, -.. , 0), verify that i. the binding restriction of a is constructively satisfied, ii. the binding restriction of 5' is not violated.</Paragraph> <Paragraph position="5"> (c) If a is a type B pronoun, antecedent candidate 5' is intrasentential, and, according to surface order, 5&quot; follows a, verify that 5' is definite.</Paragraph> <Paragraph position="6"> 2. Scoring and sorting: (a) For each remaining anaphor-candidate pair (al, 5&quot;j), determine, according to salience and parallelism heuristics, the numerical plausibility score v(ai, 5&quot;j). (b) For each anaphor a: sort candidates 7J according to decreasing plausibility v(o~, 7J). (c) Sort the anaphors a according to decreasing plausibility of their individual best antecedent candidate. null Antecedent Selection: Initialization ~ := (1,..., 1). Consider anaphors o~ in the order determined in step 2c. Suggest antecedent candidates 5&quot;(a) in the order determined in step 2b. Select 7(a) as candidate if there is no interdependency, i.e. if (a) the morphosyntactic features of c~ and 5'(o0 are still compatible, (b) for each NP (f whose coindexing with 5'(o 0 has been determined in the current invocation of the algorithm: the coindexing of a and 5 which results as a side effect when chosing 5&quot;(0) as antecedent for oL does not violate the binding principles, i.e. ~'CoL, 5) # (0,..., 0). (Here, for both occurrences, the weak predicate bp~, applies.) version holds for type A occurrences. While binding principle A constructively demands the existence of a local binder, it does not preclude further nonlocal coindexings. This prediction is in accordance with the following, intuitively acceptable example: The barberi admits that he~ shaves himself.</Paragraph> <Paragraph position="7"> During the candidate filtering phase of anaphor resolution, compliance with the binding theoretic disjoint reference rules is now verified by computing vectors F'Ca, 7)- 7 is considered a suitable candidate for a only if F'(a, 7) is not completely zero. In the course of the antecedent selection phase, the vectors of the individual decisions as well as the vectors pertaining to dynamically resulting transitive coindexings are conjoined to form a vector r ~ which characterizes the overall reading dependency: = h C/(x,v) X and Y coindexed This gives rise to a further restriction which has to be checked during the decision interdependency test step: to assure the existence of an overall reading, only choices may be combined for which r ~ does not become the zero vector.</Paragraph> </Section> <Section position="3" start_page="35" end_page="36" type="sub_section"> <SectionTitle> 5.3 An Algorithm </SectionTitle> <Paragraph position="0"> Figure 3 shows an anaphor resolution algorithm which employs the above specified method. In step 1, restrictions are applied. By determining vectors ~'(a, 7), the binding constraints are verified. In step 2, numerical preference scoring and sorting is performed, zdeg Finally, in step 3, antecedent selection takes place. Only decisions are combined which do not interdepend. The reading compatibility is verified by the stepwise computation of vector ~.11 A theoretical analysis shows that, under the assumption of a clever organization of the computation, the number of bitvector conjunctions and disjunctions (including dominance vector determination) is bounded by O(bq 2 + s), where q is the number of occurrence contributing NPs, s denotes the size of the PSF, and b stands for the maximal degree of branching due to packing and sharing. For natural language grammars, it is justified to assume that b is a (small) constant. 12 The complexity of sorting in step 2 is O(q 2 log(q)). Hence, the overall time complexity of the approach amounts to O(q2(n + log(q)) + s).</Paragraph> <Paragraph position="1"> The practical contribution of n, however, is reduced: in a reasonable implementation, the conjunction or disjunction of w bits (w = processor word length) ZdegThe optimal choice of preference factors and weights remains to be investigated; at least some of the criteria which have been investigated by (Lappin and Leass, 1994) do not immediately generalize to deficient syntax.</Paragraph> <Paragraph position="2"> lZIn some cases, it may be necessary to retract decisions. Hence, step 3 has to be supplemented with a backtracking facility.</Paragraph> <Paragraph position="3"> 12In the general case, however, b may be O(\[V\[).</Paragraph> <Paragraph position="4"> will be performed by an elementary operation.</Paragraph> </Section> <Section position="4" start_page="36" end_page="36" type="sub_section"> <SectionTitle> 5.4 Structural Disambiguation by Referential Evidence </SectionTitle> <Paragraph position="0"> Since vectors r ~ describing the reading dependency are available, any set of anaphor resolution choices may now be referred to as further evidence for structural disambiguation. PSF trees which are from now on invalid can be eliminated by pruning all packing variants whose characterizing vectors are orthogonal to ~. By this means, based on the above described framework, it becomes possible to realize a parallel processing model which accomplishes the refined version of robustness by interaction.</Paragraph> </Section> </Section> class="xml-element"></Paper>