File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/concl/05/p05-2014_concl.xml
Size: 2,045 bytes
Last Modified: 2025-10-06 13:54:43
<?xml version="1.0" standalone="yes"?> <Paper uid="P05-2014"> <Title>Dialogue Act Tagging for Instant Messaging Chat Sessions</Title> <Section position="8" start_page="83" end_page="83" type="concl"> <SectionTitle> 7 Conclusion </SectionTitle> <Paragraph position="0"> In this paper, we have highlighted some unique characteristics in IM dialogue that are not found in transcribed spoken dialogue or other forms of written dialogue such as e-mail; namely, utterance segmentation and message synchronisation. We showed the problem of unsynchronised messages can be readily solved using a simple technique utilising the typingrate and time stamps of messages. We described a method for high-accuracy dialogue act classification, which is an essential part for a deeper understanding of dialogue. In our experiments, the bi-gram model performed with the highest tagging accuracy which indicates that dialogue acts often occur as adjacency pairs. We also saw that the high tagging accuracy results obtained by the likelihood from the naive Bayes model indicated the high correlation between the actual words and dialogue acts.</Paragraph> <Paragraph position="1"> The Kappa statistics we calculated indicate that our tag set can be used reliably for annotation tasks.</Paragraph> <Paragraph position="2"> The increasing popularity of IM and automated agent-based support services is ripe with new challenges for research and development. For example, IM provides the ability for an automated agent to ask clarification questions. Appropriate dialogue modelling will enable the automated agent to reliably distinguish questions from statements. More generally, the rapidly expanding scope of online support services provides the impetus for IM dialogue systems and discourse models to be developed further.</Paragraph> <Paragraph position="3"> Our findings have demonstrated the potential for dialogue modelling for IM chat sessions, and opens the way for a comprehensive investigation of this new application area.</Paragraph> </Section> class="xml-element"></Paper>