<?xml version="1.0" standalone="yes"?>
<Paper uid="W00-1425">
  <Title>Capturing the Interaction between Aggregation and Text Planning in Two Generation Systems</Title>
  <Section position="3" start_page="186" end_page="190" type="metho">
    <SectionTitle>
2 Preferences among coherence
features
</SectionTitle>
    <Paragraph position="0"> We claim that it is the relative preferences among features rather than the absolute magnitude of each individual one that play the crucial role in the production of a coherent text. In this section we discuss the preferences among features related to text planning, based on which those for embedding can be introduced.</Paragraph>
    <Section position="1" start_page="186" end_page="187" type="sub_section">
      <SectionTitle>
2.1 Preferences for global coherence
</SectionTitle>
      <Paragraph position="0"> A semantic relation other than conjunct or disjunet is preferred to be used whenever possible because it usually conveys interesting information about domain objects and leads to a coherent text span. If a conjunct relation shares a fact with a semantic relation, the conjunct should be suppressed. For example, in 3 of Figure 1.</Paragraph>
      <Paragraph position="1"> apart from other relations, there is an amplification relation signalled by indeed and a conjunct between the last two propositions. Compared with 3, 4 is less preferred because it misses tile amplification and the center transition from the necklace to an Arts and Crafts style jewel is not so smooth, whereas 3 expresses the amplification explicitly and the conjunct implicitly.</Paragraph>
      <Paragraph position="2"> However, a semantic relation can only be used if the knowledge assumed to be shared by the hearer is introduced in the previous discourse (Mellish et al.. 1998a). \Ve assume the strategy  1. This necklace is in the Arts and Crafts style. Arts and Crafts style jewels usually have an elaborate design. They tend to have floral motifs. For instance, this necklace has floral motifs. It was designed by Jessie King. King was Scottish. She once lived in London. 2. This necklace, which was designed by Jessie King, is in the Arts and Crafts style. Arts and Crafts style jewels usually have an elaborate design. They tend to have floral motifs. For instance, this necklace has floral motifs. King was Scottish. She once lived in London. 3. The necklace is in the Arts and Crafts style. It is set with jewels in that it features cabuchon stones. Indeed, an Arts and Crafts style jewel usually uses cabuchon stones. It usually uses oval stones.</Paragraph>
      <Paragraph position="3"> 4. The necklace is in the Arts and Crafts style. It is set. with jewels in that it features cabuchon stones. An Arts and Crafts style jewel usually uses cabuchon stones and oval stones.  of (Mellish et al., 1998a) which uses a joint relation to connect every two text spans that do not have a semantic relation other than object-attribute elaboration and conjunct/disjunct in between. Although joint is not preferred when other relations are present, it is better than missing presuppositions or embedding a conjunct relation inside a semantic relation. Therefore, we have the following heuristics, where &amp;quot;A&gt;B&amp;quot; means that A is preferred over B.</Paragraph>
      <Paragraph position="4"> Heuristic 1 Preferences among features for global coherence: a semantic relation &gt; Conjunct/Disjunct &gt; Joint &gt; presuppositions not met Joint &gt; Conjunct inside a semantic relation</Paragraph>
    </Section>
    <Section position="2" start_page="187" end_page="187" type="sub_section">
      <SectionTitle>
2.2 Preferences for local coherence
</SectionTitle>
      <Paragraph position="0"> One way to achieve local coherence is to control center transitions among utterances. In Centering Theory, Rule 2 specifies preferences among center movement in a locally coherent discourse segment: sequences of continuation are preferred over sequences of retaining; which are then preferred over sequences of shifting.</Paragraph>
      <Paragraph position="1"> Brennan et el. (1987) also describe typical discourse topic movements in terms of center transitions between pairs of utterances. They argue that the order of coherence among the transitions is continuing &gt; retaining &gt; smooth shifting &gt; abrupt shifting. Instead of claiming that these are the best models, we use them simply as an example of linguistic models being used for evaluating features of text planning.</Paragraph>
      <Paragraph position="2"> A type of center transition that appears frequently in descriptive text is that the descrit)tion starts with an object, but shifts to associated objects or perspectives of that object. This is a type of abrupt shifting, but it is appropriate as long as the objects are highly associated to the original object (Schank, 1977). This phenomenon is handled in the system of (Grosz, 1977), where subparts of an object are included into a focus space as the implicit foci when the object itself is to be included.</Paragraph>
      <Paragraph position="3"> We call this center movement an associate shifting, where the center moves from a trigger entity to a closely associated entity. Our informal observation from museum descriptions shows that associate shifting is preferred by human writers to all other types of center movements except for continuation. There are two types of associate shifting: where the trigger is in the previous utterance or two entities in two adjacent utterances have the same trigger.</Paragraph>
      <Paragraph position="4"> There is no preference between them.</Paragraph>
      <Paragraph position="5"> Heuristic 2 summarises the above preferences.</Paragraph>
      <Paragraph position="6"> We admit that these are strict heuristics and that human texts are sometimes more flexible.</Paragraph>
      <Paragraph position="7">  Heuristic 2 Preferences among center transitions: null Continuation &gt; Associate shifting &gt; RetaiTIing &gt; Smooth shifting &gt; Abrupt shifting</Paragraph>
    </Section>
    <Section position="3" start_page="187" end_page="188" type="sub_section">
      <SectionTitle>
2.3 Preferences for both types of
</SectionTitle>
      <Paragraph position="0"> coherence Two propositions can be connected in different ways, e.g. through a semmxtic relation or a smooth center transition only. Since a semantic relation is always preferred, we have the following heuristic: Heuristic 3 Preferences among semantic relations and center transitions: a semantic relation &gt; Joint / Continuation</Paragraph>
    </Section>
    <Section position="4" start_page="188" end_page="190" type="sub_section">
      <SectionTitle>
2.4 Preferences for embedding Good embedding &gt; Normal embedding &gt;
</SectionTitle>
      <Paragraph position="0"> We distinguish between.a.-good,.rwrmal,and-bad Joint &gt; Bad embedding ..... =:--..~ .:-- ~ .--:.: ........</Paragraph>
      <Paragraph position="1"> embedding based on the features it bears. We do Continuation + Smooth shifting + Joint &gt; not claim that the set of features is complete.</Paragraph>
      <Paragraph position="2"> In a different context, more criteria might have to be considered.</Paragraph>
      <Paragraph position="3"> A good embedding is one satisfying all the following conditions:  1. The referring part is an indefinite, a demonstrative or a bridging description (as defined in (Poesio et al., 1997)).</Paragraph>
      <Paragraph position="4"> 2. The embedded part can be realised as an adjective or a prepositional phrase (Scott and de Souza, 1990).</Paragraph>
      <Paragraph position="5"> 3. In the resulting text, the embedded part does not lie between text spans connected by semantic parataxis and hypotaxis.</Paragraph>
      <Paragraph position="6"> 4. There is an available syntactic slot to hold  the embedded part.</Paragraph>
      <Paragraph position="7"> A good embedding is highly preferred and should be performed whenever possible. A normal embedding is one satisfying condition 1, 3 and 4 and the embedded part is a relative clause which provides additional information about the referent. Bad embeddings are all those left, for example, if there is no available syntactic slot for the embedded part.</Paragraph>
      <Paragraph position="8"> Since semantic parataxis has a higher priority than embedding (Cheng, 1998), a good embedding should be less preferred than using a conjunct relation, but it should be preferred over a center continuation for it to happen.</Paragraph>
      <Paragraph position="9"> To decide the interaction between an embedding and a center transition, we use the first two examples in Figure 1 again. The only difference between I and 2 is the position of the sentence &amp;quot;This necklace was de.signed by Jessie King&amp;quot;, which can be represented in terms of features of local coherence and embedding as follows: the last three sentences in 1: Joint + Continuation + Joint + Smooth shifting the last two sentences plus embedding in 2: Joint + Abrupt shifting + Normal embedding 1 is preferred over 2 because the center inoves more smoothly in 1. The heuristics derived from the above discussions are summarised below:  The '+' symbol can be interpreted in different ways, depending on how the features are used in an NLG system. In a traditional system, it means the coexistence of two features. In a system using numbers for planning, it can have the same meaning as the arithmetic symbol.</Paragraph>
      <Paragraph position="10"> 3 Capturing the preferences in ILEX The architecture of text planning has a great effect on aggregation possibilities. In object descriptive text generation, there lacks a central overriding communicative goal which could be decomposed in a structured way into subgoals.</Paragraph>
      <Paragraph position="11"> The main goal is to provide interesting information about the target object. There are generally only a small number of relations, mainly object-attribute elaboration and joint. For such a genre, a domain-dependent bottom-up planner (Marcu, 1997) or opportunistic planner (Mellish et al., 1998b) suits better than a domain-independent top-down planner. In these architectures, aggregation is important to text planning because it changes the order in which information is expressed. The first implementation we will describe is based on ILEX (Oberlander et al., 1998).</Paragraph>
      <Paragraph position="12"> ILEX is an adaptive hypertext generation system, providing natural language descriptions for museum objects. The bottom-up text planning is fulfilled in two steps: a content selection procedure, where a set of fact nodes with high relevance is selected from the Content Potential (following a search algorithm), and a content structuring procedure, where selected facts are reorganised to form entity-chains (based on the theory of entity-based coherence), which represent a coherent text arrangement.</Paragraph>
      <Paragraph position="13"> To make it possible for the ILEX planner to take into account aggregation, we use a revised version of Meteer's Text Structure (Meteer, 1992; Panaget, 1997) as the intermediate level of representation between text planning and sentence rcalisation to provkte abstract syntactic constraints to the planning. We call this system ILEX-TS (ILEX based on Text Structure).</Paragraph>
      <Paragraph position="14">  In ILEX-TS, abstract referring expression determination and.aggxegation are performed dur-..: ing text structuring. For each fact whose Text Structure is being built, if an NP in the fact can take modifiers, the embedding process will find a list of elaboration facts to the referent and make embedding decisions based on the constraints imposed by the NP form. The decisions include what to embed and what syntactic form the embedded part should use.</Paragraph>
      <Paragraph position="15"> Heuristic 1, 2 and 3 are followed naturally ~ by the ILEX text planner, which calculates the best RS tree and puts facts connected by the imaginary conjunct relation next to each other.</Paragraph>
      <Paragraph position="16"> It tries to feature center continuations as often as possible. When it needs to shift topic, it uses a smooth shifting.</Paragraph>
      <Paragraph position="17"> ILEX-TS has a set of embedding rules, where those rules featuring good embedding are always used first, then a rule featuring a normal embedding. Bad embedding is not allowed at all. To coordinate different types of aggregation, the algorithm checks parataxis and hypotaxis possibilities for each nucleus fact and the fact to be embedded before it applies an embedding rule. These realise most of Heuristic 4 (except for the second set). However, because the various factors are optimised in order (with no backtracking), there is no guarantee that the best overall text will be found. In addition, complex interactions between aggregation and center transition cannot be easily captured. 4 Text planning using a GA Although most heuristics can be followed in ILEX-TS, some interactions are missing, for example, 9 of Figure 1 will probably be generated. For better coordination, we adopt the text planner based on a Genetic Algorithm (GA) as described in (Mellish et al., 1998a). The task is. given a set of facts and a set of relations between facts, to produce a legal rhetoricalstrncture tree using all the facts and some relations.</Paragraph>
      <Paragraph position="18"> A fact is represented in terms of a subject, a verb and a complement (as well as a unique identifier). A relation is represented in terms of the relation name, the two facts that are connected t) 3&amp;quot; the relation and a list of precondition facts which need to have been mentioned before the relation can be used i.</Paragraph>
      <Paragraph position="19">  A genetic algorithm is suitable for such a  problem.because,:the..numher-.of.-possihle-combinations is huge and the search space is not perfectly smooth and unimodal (there can be many good combinations). Also the generation task does not require a global optimum to be found. What we need is a combination that is coherent enough for people to understand.</Paragraph>
      <Paragraph position="20"> (Mellish et al., 1998a) summarises the genetic algorithm roughly as follows:  1. Enumerate a set of random initial sequences by loosely following sequences of facts where consecutive facts mention the same entity.</Paragraph>
      <Paragraph position="21"> 2. Evaluate sequences by evaluating the rhetorical structure trees they give rise to.</Paragraph>
      <Paragraph position="22"> 3. Perform mutation and crossover on the sequences. null 4. Stop after a given number of iterations, and  return the tree for the &amp;quot;best&amp;quot; sequence. The advantage of this approach is that it provides a mechanism to integrate planning factors in the evaluation function and search for the best combinations of them. So it is an excellent framework for experimenting with the interaction between aggregation and text planning. In the algorithm, the RS trees are right-branching and are almost deterministically built from sequences of facts. Given two sequences, crossover inserts a random segment from one sequence in a random position in the other to produce two new sequences. Mutation selects a random segment of a sequence and moves it into a random position in the same sequence.</Paragraph>
      <Paragraph position="23"> To explore the whole space of aggregation.</Paragraph>
      <Paragraph position="24"> we decide not to perform aggregation on structured facts or on adjacent facts in a linear sequence because they might restrict the possibilities and even miss out good candidates. Instead, we define a third operator called embedding mutation. Suppose we have a sequence \[U1,U2,...,Ui,...,U.\], where we call each element of the sequence a unit, which can be either a fact or a list of facts or units with no depth limit. For a list, we call its very first fact the main fact, system is limited in all aspects. It does not have a real realisation component, so the parts we are less interested in are realised by canned phrases for readability.</Paragraph>
      <Paragraph position="26"> into which the remaining facts in the list are to be embedded. The embedding mutation randomly selects a unit Ui from the sequence and an entity in its main fact. It then collects all the units mentioning this entity and randomly chooses one Uk. The list containing these two units \[Ui,Uk\] represents a random embedding and will be treated as a single unit in later operations. It takes the. position of Ui to produce a new sequence \[U~,U2,...,\[Ui,Uk\],...,U,\] and all repetitions outside \[Ui,U~:\] are removed. This sequence is then evaluated and ordered in the populat ion.</Paragraph>
      <Paragraph position="27"> The probabilities of apl)lying the three operatots are: 65% for crossow'r. 30% for embedding mutation and 5% for normal umtation. This is because the first two are more likely to produce sequences bearing desired properties by either combining the good bits of two sequences or performing a reasonable amount of embedding, whereas normal mutation is entirely random 2.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>