<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-2169">
  <Title>Processing Self Corrections in a speech to speech system</Title>
  <Section position="7" start_page="1118" end_page="1118" type="concl">
    <SectionTitle>
5 Results and Further Work
</SectionTitle>
    <Paragraph position="0"> Due to the different trigger situations we performed two tests: One where we use only acoustic triggers and ~mother where the existence of a perfect word fr~gment detector is assume(1. The input were unsegmented translitera.ted utterance to exclude intluences a word 1 1 19 recognizer. We restrict the processing time on a SUN/ULTI{A 300MIIZ to 10 seconds. The parser was simulated by a word trigram. Training and testing were done on two separated parts of the German part of the Verbmobil corpus (12558 turns training / 1737 turns test).  A direct comparison to other groups is rather difficult due to very different corpora, evaluation conditions and goals. (Nakatani and Hirschberg, 1.993) suggest a acoustic/prosodic detector to identify IPs but don't discuss the problem of finding the correct segmentation in depth. Also their results are obtained on a corpus where every utterance contains at least one repair. (Shriberg, 1994) also addresses the acoustic aspects of repairs. Parsing approaches like in (Bear et al., 1992; Itindle, 1983; Core and Schubert, 1999) must be proved to work with lattices rather than transliterated text. An algorithm which is inherently capable of lattice processing is prot)osed by Heeman (Hem-nan, 1997). He redefines the word recognition problem to identify the best sequence of words, corresponding POS tags and special rel)air tags. He reports a recall rate of 81% and a precision of 83% for detection and 78%/80% tbr correction. The test settings are nearly the same as test 2. Unibrtunately, nothing is said about the processing time of his module.</Paragraph>
    <Paragraph position="1"> We have presented an approach to score potential reparandum/reparans pairs with a relative simple scope model. Our results show that repair processing with statistical methods and without deep syntactic knowledge is a promising approach at least for modification repairs. Within this fi'alnework more sophisticated scope models can be evaluated. A system integration as a filter process is described. Mapping the word lattice to a POS tag lattice is not optimal, because word inlbrmation is lost in the search tbr partial paths. We plan to implement a combined combined POS/word tagger.</Paragraph>
  </Section>
class="xml-element"></Paper>