<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-1032">
  <Title>Extracting Key Semantic Terms from Chinese Speech Query for Web Searches</Title>
  <Section position="6" start_page="2" end_page="2" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> Although research on natural language query processing and speech recognition has been carried out for many years, the combination of these two approaches to help a large population of infrequent users to &amp;quot;surf the web by voice&amp;quot; has been relatively recent. This paper outlines a divide-and-conquer approach to alleviate the effect of speech recognition error, and in extracting key CSS components for use in a standard search engine to retrieve relevant documents. The main innovative steps in our system are: (a) we use a query model to isolate CSS in speech queries; (b) we break the CSS into basic components; and (c) we employ a multi-tier approach to map the basic components to known phrases in the dictionary. The tests demonstrate that our approach is effective.</Paragraph>
    <Paragraph position="1"> The work is only the beginning. Further research can be carried out as follows. First, as most of the queries are about named entities such as the persons or organizations, we need to perform named entity analysis on the queries to better extract its structure, and in mapping to known named entities.</Paragraph>
    <Paragraph position="2"> Second, most speech recognition engine will return a list of probable words for each syllable. This could be incorporated into our framework to facilitate multi-tier mapping.</Paragraph>
  </Section>
class="xml-element"></Paper>