<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-1301">
  <Title>SPEECH-RELATED TECHNOLOGIES Where will the field go in 10 years?</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5. Proposed Actions
</SectionTitle>
    <Paragraph position="0"> Early preparations for the European Commission's 6th Framework Programme (FP6) including IST (Information Society Technologies) research are now in progress. It is premature to make predictions with any degree of certainty as to how the IST part of FP6 will shape up. Current information suggests an increased emphasis on basic research compared to the present FP5. In addition, it is possible that FP6 will include opportunities for the medium-scale research initiatives which were called for on several occasions above, i.e. large-scale &amp;quot;clusters&amp;quot; of projects all addressing the same research topic in a coordinated fashion. Finally, the current covering title for FP6 IST research is &amp;quot;ambient intelligence&amp;quot; which is one of the terms of fashion quoted in the present paper. Given the timelines and their analysis above, it does not seem to matter much which covering term is being chosen for FP6. &amp;quot;Ambient intelligence&amp;quot; is as apt as several others for FP6 and future advanced interactive systems research but, as argued in Section 3, it is far from clear if ambient intelligence requires us to focus on any particular segment of future speech-related technologies. However, the possible, increased emphasis on basic research as well as the possibility of carrying out medium-scale science in speech-related technologies are to be welcomed in the light of the argument above.</Paragraph>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
5.1 Research priorities for speech-related technologies 2000-2010
</SectionTitle>
      <Paragraph position="0"> Taking into our stride the transformations of the field of speech-related research from speech-only to interactive systems in general, and from components research to interactive systems research, the top priorities in speech-related technologies research are: * multi-speaker meeting transcription development and evaluation campaigns; * speech summarisation development and evaluation campaigns; * speech translation prototypes, generic platforms, and generic architectures.</Paragraph>
      <Paragraph position="1"> Development and evaluation campaigns are highly desirable; * conversational spoken dialogue prototypes, generic platforms, and generic architectures. Development and evaluation campaigns are highly desirable; * next-step prototypes, generic platforms, and generic architectures for intelligent multimodal information presentation; * next-step prototypes, generic platforms, and generic architectures for natural interactive systems.</Paragraph>
      <Paragraph position="2"> As soon as theoretically and practically feasible, all of the above advanced speech, multimodal and natural interactivity technologies should be developed for the web including hardware, form and contents design. The fact that some top research priorities have been mentioned above emphatically does not preclude the desirability of continuing &amp;quot;business as usual&amp;quot; in the field of speech-related research, including continued research into all of the technologies which have been mentioned earlier in the present paper. On the contrary, business as usual is actually assumed by the above top priorities list which focuses on technologies over and above business as usual. This also applies to next-step research into already deployed speech-related technologies, such as mixed initiative, task-oriented spoken dialogue systems.</Paragraph>
      <Paragraph position="3"> For basic research leading to novel concepts, theories and formalisations, the top priorities are: * basic research into human natural communicative behaviour; * a novel theory of natural communication which can replace speech acts theory and discourse theory by taking the notion of a complete communicative act as its basic notion; * research on Modality Theory in order to identify potentially useful modality combinations; * establishment of collaborative links to research into the neural basis for human natural communicative behaviour.</Paragraph>
    </Section>
    <Section position="2" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
5.2 Research organisation needed
</SectionTitle>
      <Paragraph position="0"> Medium-scale science is needed for, at least, the coordinated development of natural interactive systems prototypes, generic platforms, generic architectures, best practice in development and evaluation, and standard resources. A large, medium-scale science project with these objectives should include the porting of technologies to new languages and cultures.</Paragraph>
      <Paragraph position="1"> It is quite possible that the medium-scale science model could be applied to research into other speech-related technologies, such as speech translation technologies, conversational spoken dialogue systems, or speech technologies for ambient intelligence.</Paragraph>
      <Paragraph position="2"> For researchers in small-scale speech-related projects, in particular, the creation of a generic platforms and hardware &amp;quot;bourse&amp;quot; through contributions from European industry would be of great importance.</Paragraph>
      <Paragraph position="3"> Finally, we should stop having research programme ghettos for technologies for the disabled.</Paragraph>
    </Section>
    <Section position="3" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
5.3 Infrastructural actions needed
</SectionTitle>
      <Paragraph position="0"> In order to promote maximum uptake of the research results produced, it would be highly desirable to have funding for low-cost ways of maintaining research results for later uptake.</Paragraph>
      <Paragraph position="1"> Given the emphasis on technology development and evaluation campaigns above, Europe needs to establish an evaluation and standards agency. It is not evident to the present author that current political and industrial sensibilities would allow the US NIST to undertake to run technology development and evaluation campaigns in Europe.</Paragraph>
      <Paragraph position="2"> This having been said, there is much to be said for increasing global collaboration on many aspects of speech-related research, such as creating a coordinated global infrastructure for resources distribution.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>