<?xml version="1.0" standalone="yes"?>
<Paper uid="P92-1007">
  <Title>A Functional Approach to Generation with TAG 1</Title>
  <Section position="3" start_page="0" end_page="49" type="metho">
    <SectionTitle>
4 The task of ordering the elements of logical fonn is con-
</SectionTitle>
    <Paragraph position="0"> sidered by Mumble-86 to be part of a component wlfich is also responsible for ensuring that what is given to mmnble is actually expressible in the language (e.g., English). Tiffs component is described in (Meteer, 1991).</Paragraph>
    <Paragraph position="1"> ~Tlfis is because the logical form for an embedded question and a non-embedded question camlot be distinguished in the kind of input required by Mmnble-86 mid the main verb (wonder) is not able to pass a~ly information down to the embedded clause since it is realized after the embedded clause.</Paragraph>
    <Paragraph position="2">  and syntactic decisions are mixed does not affect the power of the generator, we argue that it does make development and maintenance of the system rather difficult. Functional decisions (e.g., that a particular item should be made prominent) and syntactic decisions (e.g., number agreement) rely on two different bodies of work which should be able to evolve independently of each other. There is no separation of these two different influences in Mumble-86.</Paragraph>
    <Paragraph position="3"> The generation process in Mumble-86 is syntax driven. From the input L-Spec an initial elementary) TAG tree is chosen. This structure s then traversed and grammar routines are initiated. At each possible attachment point during the traversal, the semantic structure (L-Spec) is consulted to see if it contains an item whose realization could be adjoined or substituted at that position. Thus the syntactic surface structure drives the processing.</Paragraph>
    <Paragraph position="4"> As a side effect of the above processing strategy, Mumble-86 creates a strictly left-to-right realization of surface structure. While this side-effect is deliberate for reasons of psychological validity, this can be problematic for generating some connectives (as is pointed out in (MeKeown &amp; E1hadad, 1991)). This is because Mumble-86 does not have access to the content of the items being conjoined at the time the connective is generated.</Paragraph>
    <Paragraph position="5"> In the remainder of this paper we describe a sentence generation system which we have developed. In some ways it is similar to Mumble-86, but there are several major differences: * The realization of the input in our system is based on systemic functional linguistics (Halliday, 1970; Halliday, 1985; Fawcett, 1980; Hudson, 1981). This is a linguistic theory which states that a generated sentence is obtained as a result of a series of functional choices which are made in a parallel fashion along several different functional domains. The choices are represented as a series of networks with traversal of the networks dependent on the given input along with several knowledge sources which encode information about how various concepts can be linguistically realized. The bulk of the work in systemic linguistics has been devoted to describing what/how functional choice affects surface form. We adopt this work from systemic linguistics, but unlike other implementations, we use a formal syntactic framework (TAG) to express the syntactic constraints.</Paragraph>
    <Paragraph position="6"> * Our method is not syntax directed, but follows a functional decomposition called for by the systemic grammar.</Paragraph>
    <Paragraph position="7"> * There is a clear separation between the functional and the syntactic aspects of sentence generation which actually allows these two aspects of generation to be developed independently. null * We do not place any constraints on the logical form input. Our methodology calls for nothing different from what is required for a standard systemic grammar (whose input is based on a typical logical form representation).</Paragraph>
    <Paragraph position="8"> * The methodology which we describe allows sentence generation to proceed in a semantic head-driven fashion (Shieber, Van Noord, Pereira ~ Moore, 1990). This is the case even for the embedded sentences discussed earlier which had to be worked &amp;quot;inside out&amp;quot; in Mumble-86.</Paragraph>
  </Section>
  <Section position="4" start_page="49" end_page="53" type="metho">
    <SectionTitle>
3 Generator Architecture
</SectionTitle>
    <Paragraph position="0"> There are many different ways of implementing a TAG-based generator. We consider the principles that we take to be common to any TAG generator and indicate how these principles have influenced our architecture. We present various aspects of our architecture and contrast them with choices that have been made in Mumble-86 and Synchronous TAG. Our approach is motivated by arguments presented in (McCoy, Vijay-Shanker Yang, i990), but the details of the processing presented there have changed significantly. Our basic processing strategy is detailed in (Yang, McCoy &amp; Vijay-Shanker, 1991); the work presented here is an extension of that strategy.</Paragraph>
    <Paragraph position="1"> In order for a TAG generator to be robust, it must have a methodology for deciphering the input and associating various pieces of the input with TAG trees. In Mumble-86 this is accomplished through dictionary look-up along with querying the input at various points during the surface structure traversal. In contrast, we use a systemic grammar traversal for this purpose. In a TAG, each elementary tree lexicalizes a predicate and contains unexpanded nodes for the required arguments. Thus any TAG based generation system should incorporate the notions of semantic head-driven generation. Our approach, based on systemic grammars, does this because the functional decomposition that results from traversal of a systemic grammar at a single rank identifies the head and establishes necessary argumentsl Thus it perfectly matches the information captured in an elementary TAG tree.</Paragraph>
    <Paragraph position="2"> Once the input has been deciphered, a TAG generator must use this to select a tree. Given that a systemic grammar is being used in our case, we must have a method for associating TAG trees with the network traversal. The traversal of a systemic grammar at a single rank establishes a set of functional choices that can be used to select a TAG tree. The selection process in any TAG-based generator can be considered as providing a classification of TAG trees on functional grounds. We make this explicit by providing a network (called the TAG network) 6 which is traversed to select a</Paragraph>
    <Paragraph position="4"/>
    <Paragraph position="6"> a decision tree whose choice points are functional features chosen in the systemic network traversal.</Paragraph>
    <Paragraph position="7"> So far we have identified how the head can be lexicalized and placed in an appropriate tree with respect to its arguments. This is accomplished by a traversal of a systemic network at one rank followed by a TAG network traversal based on the functional choices made. Of course, the arguments themselves must also be realized. This is accomplished by a recursive network (systemic followed by TAG) traversal (focused on the piece of input associated with the particular argument being realized). The recursive network traversals will also result in the realization of a TAG tree.</Paragraph>
    <Paragraph position="8"> We record information collected during a single (rank) network traversal in a data structure called a region. Thus, an initial region will be created and will record all features necessary for the selection of a tree realizing the head and argument placement. The selected tree (and other structures discussed below) will be recorded in the region. Each argument will itself be realized in a subregion which will be associated with the recursire network traversal spawned by the piece of input associated with that argument. Thus we have separate regions for each independent piece of input. This is in contrast to Mumble-86's use of the evolving surface structure in which all grammatical information is recorded.</Paragraph>
    <Paragraph position="9"> Once all arguments have been realized as elementary trees in the individual regions, the trees selected in the individual regions must be combined with the tree in the initial region. For this we use the standard TAG operations of adjoining and substitution.</Paragraph>
    <Paragraph position="10"> Essentially, our generation methodology consists of two phases:  sible for spawning the creation of subregions in which the arguments (and modifiers) are realized.</Paragraph>
    <Paragraph position="11"> 2. The ascent process - where the trees created in the individual subregions are combined with the tree in the mother region resuiting in the final realization of the whole.</Paragraph>
    <Paragraph position="12"> In our system the systemic network traversal basically replaces the dictionary look-up phase found in Mumble-867 which translates the input L-Spec into surface structure. In addition, our system does not walk a surface structure (i.e., the actual tree chosen). In Mumble-86 the surface structure walk spawned grammar routines and caused additional pieces of the L-Spec to be translated into surface structure. Our methodology relies on the systemic network traversal to spawn realizations of the decomposed subpieces. The syntactic aspects of the grammar routines are now incorporated into our TAG network and grammar.</Paragraph>
    <Paragraph position="13"> Thus our methodology keeps a clearer separation between functional and syntactic aspects of the generation process.</Paragraph>
    <Paragraph position="14"> The processing in our system will be explained with an example. Consider the simplified input given in Figure 1. s See (Yang, McCoy &amp; Vijay-Shanker, 1991) for a more detailed description of the processing.</Paragraph>
    <Paragraph position="15"> ;'The systenxic grammar also replaces the grammar routines of Mmnble-86 responsible for recording contextual information for subsequent translations. In addition, the part of the dictionary look-up concerned with syntactic realization (i.e., the actual tree chosen) is handled by our TAG component.</Paragraph>
    <Paragraph position="16"> STiffs input is simplified in that it is basically a standard logical form input with lexicM items specified. In general the input is a set of features wlffch drive the traversal of the ftmctional systemic networks.</Paragraph>
    <Section position="1" start_page="51" end_page="51" type="sub_section">
      <SectionTitle>
3.1 The Descent Process
</SectionTitle>
      <Paragraph position="0"> The input given (along with other knowledge sources traditionally associated with a systemic network) will be used to drive the traversal of a functional systemic network. The purpose of this traversal is two fold: (1) to identify the head/argument structure of the sentence to be realized, and (2) to identify a set of functional features which can be used to choose a tree which appropriately realizes the head/argument structure.</Paragraph>
      <Paragraph position="1"> Traditionally a systemic network consists of a number of networks of functional choices which are traversed in parallel. Each network considers choices along one functional domain. One such network is the mood network which is responsible for, among other things, determining what kind of speech act should be generated for the top-level element. This network must notice, for example, that the speech-act specified is wh-questioning, but that the item being questioned is not one of the arguments to the top level process. Thus a standard declarative form should be chosen for the realization of this top level element.</Paragraph>
      <Paragraph position="2"> Standard implementations of systemic grammar (Davey, 1978; Mann &amp; Matthiessen, 1985; Patten, 1988; Fawcett, 1990), upon traversal of the mood network to this point, would evaluate a set of realization operations which manipulate an eventual surface string. For instance, upon identifying that a declarative form is needed, the subject would be ordered before the finite. We argue in (McCoy, Vijay-Shanker &amp; Yang, 1990) that it is more practical to replace the use of such realization operators with a more formal grammatical system (and that the use of such a system is perfectly consistent with the tenets of systemic linguistics). Thus during the network traversal, our system simply collects the chosen features and these are used to drive the traversal of a TAG network whose traversal results in the selection of a tree.</Paragraph>
      <Paragraph position="3"> At the same time the mood network is traversed, so would be other networks. The transitivity network is concerned with identifying the head argument structure of the item being realized. In</Paragraph>
      <Paragraph position="5"> this case, it would consider the fact that the item to be realized has a &amp;quot;process&amp;quot; which is mental.</Paragraph>
      <Paragraph position="6"> This identification results in the expectation of two arguments - an actor (doing the mental process) and a phenomenon (that thing the process is about). Each of these identified arguments must be realized individually. This is accomplished via the pveselect operation2 This operation causes a recursive network traversal (whose results are recorded in a subregion) to be done focused on the input for the identified sub-element.</Paragraph>
      <Paragraph position="7"> The features collected during the functional systemic network traversal are used to drive the traversal of the TAG network which results in the selection of a tree realizing the indicated features.</Paragraph>
      <Paragraph position="8"> Features such as that the process is mental and that the speech act is declarative would cause the selection of a tree for the mother region such as the tree in Figure 2.</Paragraph>
      <Paragraph position="9"> Similar processing would then take place in the two subregions, each eventually resulting in the trees such as those shown in Figures 3 and 4.</Paragraph>
    </Section>
    <Section position="2" start_page="51" end_page="53" type="sub_section">
      <SectionTitle>
3.2 The Ascent Process
</SectionTitle>
      <Paragraph position="0"> In a TAG generator, after the input has been decomposed and elementary trees associated with each subpiece of the input, the chosen trees must be put together. Therefore, every TAG generator must provide a means to determine where 9 From the realization operations used in systemic grmnmars (particularly Nigel), we need only the preselect and the conflate operations because all structure building operations are incorporated into TAG. The conflation operation is used to map functional features (e.g., agent, phenomenon) into granunatical functions (e.g., subject, complement). Note that in the networks from systemic grammars, we take ouly the functional part and thus avoid having choice points that exist for purely syntactic reasons.  the substitution or adjunction must take place. In order to do this, with each tree there must be a mapping of grammatical functions to nodes in the tree. In our case, we associate a mapping table with each tree. For instance, the mapping table associated with the tree shown in Figure 2 would indicate that the phenomenon (which would have been conflated with complement) is associated with the node labeled nl in the tree. In the simplest case the tree which realizes the phenomenon would be substituted at the node labeled nl in the tree in the mother region.</Paragraph>
      <Paragraph position="1"> A data structure similar to a mapping table is used by the other TAG generators as well. In synchronous TAG the mapping table corresponds to the explicit node for node mapping between elementary logical form and syntactic trees. The mapping table in Mumble-86 is implicit in the schemas which create the surface structure tree (during the dictionary look-up phase) since they place L-spec elements in the appropriate place in the surface structure they create.</Paragraph>
      <Paragraph position="2"> A more complex case arises when an argument node is a footnode of an auxiliary tree. Suppose an auxiliary tree, fl, was chosen in a region and a tree, 7, was chosen in a subregion to realize the argument specified by the footnode of ft.</Paragraph>
      <Paragraph position="3"> Rather than substituting 7 in/3, fl is adjoined into a node in 7- This node is the node in 7 that heads the subtree realizing the function specified for the subregion. For this reason, each tree in a region also has associated with it a pointer we call an fr-node which points to the node heading this subtree (functional root). In Regions rl and r2 the functional root is also the root of the tree. Notice in Region r3 that the functional root is the embedded S node. This fr-node is chosen because the tree chosen in the region is a wh-question tree due to the fact that (according to the input) the phenomenon is being questioned. There is nothing in the phenomenon itself, however, that specifies that  its speech-act should be wh-questioning. Thus the portion of the tree under the embedded S node captures the predicate argument structure which realizes the phenomenon as is specified in the input. If it were the case that the phenomenon was specified to be a wh-question (as in &amp;quot;Mary wondered who hit John&amp;quot;) then the root node would be chosen as the fr-node. The fr-node comes into play when the trees in the individual regions are combined via adjunction during the ascent process.</Paragraph>
      <Paragraph position="4"> Other TAG generators have analogues to our fr-node. In synchronous TAG it is implicit in the mapping between the nodes in the two kinds of trees. In Mumble-86, it is the attachment points on surface structure. The point is that if trees might be adjoined into, any TAG generator must specify where adjoining might take place and this specification depends (at least in part) on the functional content that the tree is intended to capture. Going back to our example, in combining trees in the subregions with the tree chosen in the initial region rl, the agent tree would be combined with the tree in region rl using straight substitution. The location of the substitution would be determined by the address given for the agent in the mapping table for the tree in region rl.</Paragraph>
      <Paragraph position="5"> The mapping table also indicates that the phenomenon should be placed at nl in the tree in Figure 2. Notice, however, that nl is the foot node. This is an indication to the processor that the final tree in region rl should result from adjoining the tree in rl into the tree in the subregion r3 (Figure 4). The place of adjoining is specified by the fr-node in the phenomenon tree in region</Paragraph>
    </Section>
  </Section>
  <Section position="5" start_page="53" end_page="53" type="metho">
    <SectionTitle>
4 Passing Features
</SectionTitle>
    <Paragraph position="0"> So far we have established that any TAG-based generator, once an elementary tree has been chosen, would need to realize the arguments of the predicate by recursively calling the same procedure. The resulting trees chosen would be combined with the original elementary tree at the appropriate place by substitution and adjunction. In this recursive process, we have indicated the need for only functional information to be passed down from the mother region to the subregions (at the very least, in the form of the functional input associated with the piece being realized in the region). We now consider an example where syntactic information must be passed down as well.</Paragraph>
    <Paragraph position="1"> Consider the generation of a sentence such as &amp;quot;John tried to win&amp;quot;. The standard structure for this sentence is given in Figure 6. The problem is that in TAG this tree must be derived from the combination of two separate sentential trees: one headed by the verb &amp;quot;tried&amp;quot; and the other by the verb &amp;quot;win&amp;quot;. However we must capture the constraint that the subject of the &amp;quot;win&amp;quot; tree is John (which is the same as the subject of the &amp;quot;tried&amp;quot; (Yang, 1991). It is inserted in the region rl as a result of a feature disparity on the nodes of the tree resulting from the adjoining operation just described. The same disparity would not occur in indirect questions (e.g., &amp;quot;I wonder who kit Jolm&amp;quot; ).</Paragraph>
    <Paragraph position="2"> tree) but that it is realized only as a (null) pro.</Paragraph>
    <Paragraph position="3"> Note that this constraint cannot be localized in TAG but cuts across two elementary trees.</Paragraph>
    <Paragraph position="4"> While generating this sentence, when we choose the &amp;quot;tried&amp;quot; tree in the mother region, we must pass down the information that among the trees associated with win, the one with &amp;quot;pro&amp;quot; in the subject position must be chosen. Notice that this is a purely syntactic constraint based on the choice of the verb &amp;quot;try&amp;quot;. The choosing of this tree has ramifications on both the functional network traversal (since the agent of &amp;quot;win&amp;quot; should not be expanded) and the TAG network traversal.</Paragraph>
    <Paragraph position="5"> In addition, any syntactic constraint that is placed on the arguments (perhaps by the choice of the head) must be passed down to the subregion to influence the realization of the arguments. In general, the passed down features may influence either the functional or the TAG network traversal (see Figure 7). Such passing of syntactic and functional features must occur in any TAG generator where the realization of the head is done prior to the realization of its arguments.</Paragraph>
  </Section>
class="xml-element"></Paper>