File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/intro/03/p03-2005_intro.xml
Size: 2,057 bytes
Last Modified: 2025-10-06 14:01:50
<?xml version="1.0" standalone="yes"?> <Paper uid="P03-2005"> <Title>An Adaptive Approach to Collecting Multimodal Input</Title> <Section position="3" start_page="0" end_page="0" type="intro"> <SectionTitle> 2 Motivation </SectionTitle> <Paragraph position="0"> When providing composite multimodal input, i.e.</Paragraph> <Paragraph position="1"> input that needs to be interpreted or combined together for proper understanding, the user has flexibility in the timing of those multimodal inputs.</Paragraph> <Paragraph position="2"> Considering two inputs at a time, the user can input them either sequentially or simultaneously. A multimodal input may consist of more than two inputs, leading to a large number of composite schemes.</Paragraph> <Paragraph position="3"> MMIF needs to deal with these complex schemes and determine a suitable time when it is most unlikely to receive any further input and indicate the end of a user turn.</Paragraph> <Paragraph position="4"> The determination of the end of a user turn becomes a problem because of the following two conflicting requirements: 1. For naturalness, the user should not be constrained by pre-defined interaction requirements, e.g. to speak within a specified time after touching the display. To allow this flexibility in the sequential interaction metaphor, the user can provide coordinated multimodal input anytime after providing input in some modality. Also each modality has a unique processing time requirement due to differing resource needs and capture times e.g. spoken input takes longer compared with touch. The MMIF needs to consider such delays before sending information to a dialogue manager (DM). These requirements tend to increase the time to wait for further information from input modalities.</Paragraph> <Paragraph position="5"> 2. Users would expect the system to respond as soon as they complete their input. Thus, the fusion module should take as little time as possible before sending the integrated information to the dialogue manager.</Paragraph> </Section> class="xml-element"></Paper>