<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-2001">
  <Title>Multilingual Speech Recognition for Information Retrieval in Indian context Udhyakumar.N, Swaminathan.R and Ramakrishnan.S.K</Title>
  <Section position="10" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> This work presents the recent results in building a full-fledged multilingual speech recognizer for our ongoing project 'Telephone-based spoken language information retrieval system for Indian languages'. Techniques like CART based LTS, language identification using bi-grams and accented English recognition by native language bootstrapping have been experimented.</Paragraph>
    <Paragraph position="1"> Significant amount of research remains in handling spontaneous speech effects and nonverbal sounds, which are common in real world data. We have planned to explore language modeling and adaptive signal processing techniques to address these problems (Acero.A 1993). Use of model interpolation and weighted finite state transducers (Karen Livescu 1999) are presently analyzed to improve the system performance on accented English. From our experience we understood the importance of the acronym 'while there is no data like more data, there is also no data like real data'. Hence we have started data collection to carry out next phase of experiments.</Paragraph>
  </Section>
class="xml-element"></Paper>