<?xml version="1.0" standalone="yes"?>
<Paper uid="W96-0404">
  <Title>Approximate Generation from Non-Hierarchical Representations</Title>
  <Section position="4" start_page="32" end_page="33" type="metho">
    <SectionTitle>
3 D-Tree Grammars
</SectionTitle>
    <Paragraph position="0"> Our generator uses a particular syntactic theory--D-Tree Grammar (DIG) which we briefly introduce because the generation strategy is influenced by the linguistic structures and the operations on them.</Paragraph>
    <Paragraph position="1"> D-Tree Grammar (DTG) [16] is a new grammar formalism which arises from work on Tree-Adjoining Grammars (TAG) [7]. In the context of generation, TAGS have been used in a number of systems MUMBLE [10], SPOKESMAN [11], Wm [27], the system reported in [9], the first version of PROTECTOR [12], and recently SPUD (by Stone &amp; Doran). In the area of grammar development TAG has been the basis of one of the largest grammars developed for English [4].</Paragraph>
    <Paragraph position="2"> Unlike TAGs, DTGs provide a uniform treatment of complementation and modification at the syntactic level. DTGs are seen as attractive for generation because a close match between semantic and syntactic operations leads to simplifications in the overall generation architecture. DTGS try to overcome the problems associated with TAGS while remaining faithful to what is seen as the key advantages of TAGs [7]: the extended domain of locality over which syntactic dependencies are stated and function argument structure is captured. null DTG assumes the existence of elementary structures and uses two operations to form larger structures from smaller ones. The elementary structures are tree descriptions 3 which are trees in which nodes are linked with two types of links: domination links (d-links) and immediate domination links (i-links) expressing (reflexive) domination and immediate domination relations 3called d-trees hence the name of the formalism.</Paragraph>
    <Paragraph position="3"> between nodes. Graphically we will use a dashed line to indicate a d-link (see Figure 2). D-trees allow us to view the operations for composing trees as monotonic. The two combination operations that DTG uses are subsertion and sister- null Subsertion. When a d-tree a is subserted into another d-tree fl, a component 4 of a is substituted at a frontier nonterminal node (a substitution node) of j3 and all components of a that are above the substituted component are inserted into d-links above the substituted node or placed above the root node of ft. It is possible for components above the substituted node to drift arbitrarily far up the d-tree and distribute themselves within domination links, or above the root, in any way that is compatible with the domination relationships present in the substituted d-tree. In order to constrain the way in which the non-substituted components can be interspersed DTG uses subsertion-insertion constraints which explicitly specify what components from what trees can appear within a certain d-links. Subsertion as it is defined as a non-deterministic operation. Subsertion can model both adjunction and substitution in TAG .</Paragraph>
    <Paragraph position="4">  posed d-tree 7 results from the addition to j3 of v~ as a new leftmost or rightmost sub-d-tree below 7/. Sister-adjunction involves the addition of exactly one new immediate domination link. In addition several sister-adjunctions can occur at the same node. Sister-adjoining constraints associated with nodes in the d-trees specify which other d-trees can be sister-adjoined at this node and whether they will be right- or left-sisteradjoined. null For more details on DTGS see \[16\].</Paragraph>
  </Section>
  <Section position="5" start_page="33" end_page="34" type="metho">
    <SectionTitle>
4 Knowledge Sources
</SectionTitle>
    <Paragraph position="0"> The generator assumes it is given as input an input semantics (InputSem) and 'boundary' constraints for the semantics of the generated sentence (BuiltSem which in general is different from InputSemh). The boundary constraints are two graphs (UpperSem and LowerSem) which convey the notion of the least and the most that should be expressed. So we want BuiltSem to satisfy: LowerSern &lt; BuiltSem &lt;_ UpperSern. C/ If the generator happens to introduce more semantic information by choosing a particular expression, LowerSem is the place where such additions can be checked for consistency. Such constraints on BuiltSem are useful because in general InputSem and BuiltSem can happen to be incomparable (neither one subsumes the other). In a practical scenario LowerSem can be the knowledge base to which the generator has access minus any contentious bits. UpperSem can be the minimum information that necessarily has to be conveyed in order for the generator to achieve the initial communicative intentions.</Paragraph>
    <Paragraph position="1"> The goal of the generator is to produce a sentence whose corresponding semantics is as close as possible to the input semantics, i.e., the realisation adds as little as possible extra material and misses as little as possible of the original input. In generation similar constraints have been used in the generation of referring expressions where the expressions should not be too general  by G2. We consider UpperSem to be a generalisation of BuiltSem and LowerSem a specialisation of BuiltSem (in terms of the conceptual graphs that represent them). so that discriminatory power is not lost and not too specific so that the referring expression is in a sense minimal. Our model is a generalisation of the paradigm presented in \[17\] where issues of mismatch in lexical choice are discussed. We return to how UpperSem and LowerSem are actually used in Section 7.</Paragraph>
    <Section position="1" start_page="33" end_page="34" type="sub_section">
      <SectionTitle>
4.1 Mapping rules
</SectionTitle>
      <Paragraph position="0"> Mapping rules state how the semantics is related to the syntactic representation. We do not impose any intrinsic directionality on the mapping rules and view them as declarative statements.</Paragraph>
      <Paragraph position="1"> In our generator a mapping rule is represented as a d-tree in which certain nodes are annotated with semantic information. Mapping rules are a mixed syntactic-semantic representation.</Paragraph>
      <Paragraph position="2"> The nodes in the syntactic structure will be feature structures and we use unification to combine two syntactic nodes. The semantic annotations of the syntactic nodes are either conceptual graphs or instructions indicating how to compute the semantics of the syntactic node from the semantics of the daughter syntactic nodes.</Paragraph>
      <Paragraph position="3"> Graphically we use dotted lines to show the coreference between graphs (or concepts). Each graph appearing in the rule has a single node (&amp;quot;the semantic head&amp;quot;) which acts as a root (indicated by an arrow in Figure 4). This hierarchical structure is imposed by the rule, and is not part of the semantic input. Every mapping rule has associated applicability semantics which is used to license its application. The applicability semantics can be viewed as an evaluation of the semantic instruction associated with the top syntactic node in the tree description.</Paragraph>
      <Paragraph position="4"> Figure 4 shows an example of a mapping rule.</Paragraph>
      <Paragraph position="5"> The applicability semantics of this mapping rule is: I AN'MATE ACT,ON If this structure matches part of the input semantics (we explain more precisely what we mean by matching later on) then this rule can be triggered (if it is syntactically appropriate--see Section 5). The internal generation goals (shaded areas) express the following: (1) generate \[ACTION\[ as a verb and subsert (substitute,attach) the verb's syntactic structure at the Vo node; (2) generate \[ANIMATE\] as a noun phrase and subsert the newly built structure at NPO; and (3) generate I EI~ITITY\[ aS another noun phrase and subsert the newly built struc- null notated d-trees) and they are incorporated in the mixed structure corresponding to the current status of the generated sentence.</Paragraph>
    </Section>
  </Section>
  <Section position="6" start_page="34" end_page="35" type="metho">
    <SectionTitle>
5 Sentence Generation
</SectionTitle>
    <Paragraph position="0"> In this section we informally describe the generation algorithm. In Figure 5 and later in Figure 8, which illustrate some semantic aspects of the processing, we use a diagrammatic notation to describe semantic structures which are actually encoded using conceptual graphs.</Paragraph>
    <Paragraph position="1"> The input to the generator is InputSem, LowerSem, UpperSem and a mixed structure, Partial, which contains a syntactic part (usually just one node but possibly something more complex) and a semantic part which takes the form of semantic annotations on the syntactic nodes in the syntactic part. Initially Partial represents the syntactic-semantic correspondences which are imposed on the generator. 7 It has the format of a mixed structure like the representation used to express mapping rules (Figure 4). Later during the generation Partial is enriched and at any stage of processing it represents the current syntactic-semantic correspondences.</Paragraph>
    <Paragraph position="2"> We have augmented the DTG formalism so 7In dialogue and question answering, for example, the syntactic form of the generated sentence may be constrained.</Paragraph>
    <Paragraph position="3"> that the semantic structures associated with syntactic nodes will be updated appropriately during the subsertion and sister-adjunction operations. The stages of generation are: (1) building an initial skeletal structure; (2) attempting to consume as much as possible of the semantics uncovered in the previous stage; and (3) converting the partial syntactic structure into a complete syntactic tree.</Paragraph>
    <Section position="1" start_page="34" end_page="35" type="sub_section">
      <SectionTitle>
5.1 Building a skeletal structure
</SectionTitle>
      <Paragraph position="0"> Generation starts by first trying to find a mapping rule whose semantic structure matches s part of the initial graph and whose syntactic structure is compatible with the goal syntax (the syntactic part of Partial). If the initial goal has a more elaborate syntactic structure and requires parts of the semantics to be expressed as certain syntactic structures this has to be respected by the mapping rule. Such an initial mapping rule will have a syntactic structure that will provide the skeleton syntax for the sentence.</Paragraph>
      <Paragraph position="1"> If Lexicalised DTGiS used as the base syntactic formalism at this stage the mapping rule will introduce the head of the sentence structure the main verb. If the rule has internal generation goals then these are explored recursively (possibly via an agenda--we will ignore here the Svia the maximal join operation. Also note that the arcs to/from the conceptual relations do not reflect any directionality of the processing--they can be 'traversed'/accessed from any of the nodes they connect.  issue of the order in which internal generation goals are executed). Because of the minimality of the mapping rule, the syntactic structure that is produced by this initial stage is very basic--for ex:mple only obligatory complements are considered. Any mapping rule can introduce additional semantics and such additions are checked against the lower semantic bound. When applying a mapping rule the generator keeps track of how much of the initial semantic structure has been covered/consumed. Thus at the point when all internal generation goals of the first (skeletal) mapping rule have been exhausted the generator knows how much of the initial graph remains to be expressed.</Paragraph>
    </Section>
    <Section position="2" start_page="35" end_page="35" type="sub_section">
      <SectionTitle>
5.2 Covering the remaining semantics
</SectionTitle>
      <Paragraph position="0"> In the second stage the generator aims to find mapping rules in order to cover most of the remaining semantics (see Figure 5) . The choice of mapping rules is influenced by the following criteria: Connectivity: The semantics of the mapping rule has to match (cover) part of the covered semantics and part of the remaining semantics. null Integration: It should be possible to incorporate the semantics of the mapping rule into the semantics of the current structure being built by the generator.</Paragraph>
      <Paragraph position="1"> Realisability: It should be possible to incorporate the partial syntactic structure of the mapping rule into the current syntactic structure being built by the generator.</Paragraph>
      <Paragraph position="2"> Note that the connectivity condition restricts the choice of mapping rules so that a rule that matches part of the remaining semantics and the extra semantics added by previous mapping rules cannot be chosen (e.g., the &amp;quot;bad mapping&amp;quot; in Figure 5). While in the stage of fleshing out the skeleton sentence structure (Section 5.1) the syntactic integration involves subsertion, in the stage of covering the remaining semantics it is sister-adjunction that is used. When incorporating semantic structures the semantic head has to be preserved--for example when sister-adjoining the d-tree for an adverbial construction the semantic head of the top syntactic node has to be the same as the semantic head of the node at which sister-adjunction is done. This explicit marking of the semantic head concepts differs from \[20\] where the semantic head is a PROLOG term with exactly the same structure as the input semantics.</Paragraph>
    </Section>
    <Section position="3" start_page="35" end_page="35" type="sub_section">
      <SectionTitle>
5.3 Completing a derivation
</SectionTitle>
      <Paragraph position="0"> In the preceding stages of building the skeletal sentence structure and covering the remaining semantics, the generator is mainly concerned with consuming the initial semantic structure.</Paragraph>
      <Paragraph position="1"> In those processes, parts of the semantics are mapped onto partial syntactic structures which are integrated and the result is still a partial syntactic structure. That is why a final step of &amp;quot;closing off&amp;quot; the derivation is needed. The generator tries to convert the partial syntactic structure into a complete syntactic tree. A morphological post-processor reads the leaves of the final syntactic tree and inflects the words.</Paragraph>
    </Section>
  </Section>
  <Section position="7" start_page="35" end_page="37" type="metho">
    <SectionTitle>
6 Example
</SectionTitle>
    <Paragraph position="0"> In this section we illustrate how the algorithm works by means of a simple example. Suppose  number of ways: Fred limped quickly, Fred hurried with a limp, Fred's limping was quick, The quickness of Fred's limping ..., etc. Here we show how the first paraphrase is generated.</Paragraph>
    <Paragraph position="1"> In the stage of building the skeletal structure the mapping rule (i) in Figure 6 is used. Its internal generation goals are to realise the instantiation of \[ ACTION \] (which is \[ MOVEMENT as a verb and similarly\[ PERSON:FRED f as a noun phrase. The generation of the subject noun phrase is not discussed here. The main verb is generated using the terminal mapping rule 9 (iii) in Figure 6. ldeg The skeletal structure thus generated is Fred limp(ed). (see (i) in Figure 7). An interesting point is that although the internal generation goal for the verb referred only to the concept \[MOVEMENT\] in the initial semantics, all of the information suggested by the terminal mapping rule (iii) in Figure 6 is consumed. We will say more about how this is done in Section 7.</Paragraph>
    <Paragraph position="2"> At this stage the only concept that remains to be consumed is \[~K~. This is done in the stage of covering the remaining semantics when the mapping rule (ii) is used. This rule has an internal generation goal to generate the instantiation Of\[MANNER\] as an adverb, which yields quickly. The structure suggested by this rule has to be integrated in the skeletal structure.</Paragraph>
    <Paragraph position="3"> degTerminal mapping rules are mapping rules which have no internal generation goals and in which all terminal nodes of the syntactic structure are labelled with terminal symbols (lexemes).</Paragraph>
    <Paragraph position="4">  present in the initial trees.</Paragraph>
    <Paragraph position="5"> On the syntactic side this is done using sisteradjunction. The final mixed syntactic-semantic structure is shown on the right in Figure 7. In the syntactic part of this structure we have no domination links. Also all of the input semantics has been consumed. The semantic annotations of the S and VP nodes are instructions about how the graphs/concepts of their daughters are to be combined. If we evaluate in a bottom up fashion the semantics of the S node, we will get the same result as the input semantics in  the result is Fred limped quickly. An alternative paraphrase like Fred hurried with a limp ll can be generated using a lexical mapping rule for the verb hurry which groups IMOVEMENTI and \[~ together and a another mapping rule expressing \[LIMPING\] as a PP. To get both paraphrases would be hard for generators relying on hierarchical representations.</Paragraph>
    <Paragraph position="6"> 7 Matching the applicability semantics of mapping rules Matching of the applicability semantics of mapping rules against other semantic structures occurs in the following cases: when looking for a skeletal structure; when exploring an internal generation goal; and when looking for mapping rules in the phase of covering the remaining semantics. During the exploration of internal generation goals the applicability semantics of a mapping rule is matched against the semantics of an internal generation goal. We assume that 11 Our example is based on Iordanskaja e~ al.'s notion of maximal reductions of a semantic net (see \[6, page 300\]). It is also similar to the example in \[14\].</Paragraph>
  </Section>
  <Section position="8" start_page="37" end_page="37" type="metho">
    <SectionTitle>
IN LOWER SEM. BOUND
SEMANTICS OF THE I ~ .~ -, ,,~.~'~.~
GENERATION GOAL ~ ~&amp;quot;'~'~&amp;quot;'~ I X
\
APPLICABILITY SEMANTICS
OF NEW MAPPING RULE INITIAL GRA
</SectionTitle>
    <Paragraph position="0"> rule that is more specialised than the goal semantics (additional concepts/relations, further type instantiation, etc.) must be within the lower semantic bound (LowerSem). If this additional information is within the input semantics, then information can propagate from the input semantics to the mapping rule (the shaded area 2 in Figure 8). If the mapping rule's semantic additions are merely in LowerSem, then information cannot flow from LowerSem to the mapping rule (area 1 in Figure 8).</Paragraph>
    <Paragraph position="1"> Similar conditions hold when in the phase of covering the remaining ~emantics the applicability semantics of a mapping rule is matched against the initial semantics. This way of matching allows the generator to convey only the information in the original semantics and what the language forces one to convey even though more information might be known about the particular situation.</Paragraph>
    <Paragraph position="2"> In the same spirit after the generator has consumed/expressed a concept in the input semantics the system checks that the lexical semantics of the generated word is more specific than the corresponding concept (if there is one) in the upper semantic bound.</Paragraph>
  </Section>
  <Section position="9" start_page="37" end_page="38" type="metho">
    <SectionTitle>
8 Implementation
</SectionTitle>
    <Paragraph position="0"> We have developed a sentence generator called PROTECTOR (approximate PROduction of TExts from Conceptual graphs in a declaraTive framewORk). PROTECTOR is implemented in LIFE \[1\]. The syntactic coverage of the generator is influenced by the XTAG system (the first version of PROTECTOR in fact used TAGS). By using DTGs we can use most of the analysis of XTAG while the generation algorithm is simpler.</Paragraph>
    <Paragraph position="1"> W~ are in a position to express subparts of the input semantics as different syntactic categories as appropriate for the current generation goxl (e.g., VPs and nominalisations). The syntactic  coverage of PROTECTOR includes: intransitive, transitive, and ditransitive verbs, topicalisation, verb particles, passive, sentential complements, control constructions, relative clauses, nominalisations and a variety of idioms. On backtracking PROTECTOR returns all solutions. We are also looking at the advantages that our approach offers for multilingual generation.</Paragraph>
  </Section>
class="xml-element"></Paper>