File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/intro/94/w94-0302_intro.xml
Size: 4,603 bytes
Last Modified: 2025-10-06 14:05:46
<?xml version="1.0" standalone="yes"?> <Paper uid="W94-0302"> <Title>DPOCL: A Principled Approach to Discourse Planning</Title> <Section position="3" start_page="0" end_page="13" type="intro"> <SectionTitle> 2 Representation in Discourse Plans </SectionTitle> <Paragraph position="0"> Previous approaches have viewed the discourse planner as a means to producing a specification of a discourse that can be given to a text realization system in order to produce a series of sentences in a natural language.</Paragraph> <Paragraph position="1"> Recent work has shown that plans play a much larger role in agent interaction \[16\]. In particular, the structure of discourse plans plays a role in the comprehension of the discourse \[6, 11, 16\] and contributes to the nature of subsequent communication \[15, 24\].</Paragraph> <Section position="1" start_page="0" end_page="13" type="sub_section"> <SectionTitle> 2.1 Representing Intentional Structure </SectionTitle> <Paragraph position="0"> As has been noted \[15, 16, 24\], a precise definition of intention in discourse plans is crucial for enabling systems to respond appropiiately to failures of their communicative actions. When a hearer reveals that an in-</Paragraph> </Section> <Section position="2" start_page="13" end_page="13" type="sub_section"> <SectionTitle> 7th International Generation Workshop * Kennebunkport, Maine * June 21-24, 1994 </SectionTitle> <Paragraph position="0"> tended effect of a previous discourse did not succeed, the speaker should re-try to achieve that effect. If, however, the effect that failed was not an intended effect, the speaker need not generate an alternative response to achieve it. Alternatively, if the effect that failed was intended, but served only as a precondition of an action whose intended effects succeeded despite the failure, then again the speaker may chose not to respond. Clearly, differentiating between intended and unintended effects of discourse actions is critical for generating appropriate responses.</Paragraph> <Paragraph position="1"> As Maybury has pointed out \[13\], a realistic description of communicative action requires a representation that allows individual actions to have more than one effect on the mental state of the hearer. In particular, abstract communicative actions need to be described in a way that represents at least some of the effects of the steps in their subplans. Allowing action descriptions that have multiple effects complicates the definition of intentional structure. The reason for inserting a step in a plan is to establish some intended condition(s). However, when steps have multiple effects, it may be the case that only some of these effects are used to establish intended conditions in any given context. Any effects of a step that do not play a role in establishing such conditions in a given plan are considered side effects.</Paragraph> <Paragraph position="2"> In Section 6.1, we formally define intention in the DPOCL framework, and show how intended effects are distinguished from side effects when action descriptions may have multiple effects.</Paragraph> </Section> <Section position="3" start_page="13" end_page="13" type="sub_section"> <SectionTitle> 2.2 Representing Informational Structure </SectionTitle> <Paragraph position="0"> Just as the structure of a discourse reflects the intentions of the speaker, that structure also reflects the way in which domain content is used to achieve intended effects.</Paragraph> <Paragraph position="1"> This informational structure captures relationships that hold between objects in the domain of discourse. In an explanation, for example, one utterance may describe an event that can be presumed to be the cause of another event described in the subsequent utterance.</Paragraph> <Paragraph position="2"> Clearly intention and information are closely related.</Paragraph> <Paragraph position="3"> An important component of an agent's linguistic capsbility is the knowledge of what types of information can be used to achieve communicative intentions. Hearers may be able to determine what the speaker is trying to do because of what the hearer knows about the world or what she knows about what the speaker believes about the world. Alternatively, the hearer may be able to figure out what the speaker believes about the world by recognizing what the speaker is trying to do in the discourse \[16\]. In Section 6.1, we describe how decomposition operators in DPOCL capture the relationship between intentional and informational structure.</Paragraph> </Section> </Section> class="xml-element"></Paper>