<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-2103">
  <Title>Granularity Effects in Tense Translation</Title>
  <Section position="7" start_page="716" end_page="717" type="evalu">
    <SectionTitle>
8 Results
</SectionTitle>
    <Paragraph position="0"> We tested the system on a data base of 13,625 pairs of Gernlan VI~I\]q with their English translations, containing 12,036 tensed verbs. All the d~tl;a were in the appointment scheduling domain which is investigated in Verbmobil.</Paragraph>
    <Paragraph position="1"> They wore transliter~ted and syntactically annotated 1)3; hand. rlk'anslation was perforined by the Verbmol)il transfer component (Dorna and Emele, 1996). 2,758 tensed verbs were modified by time. adverbiMs, 1,373 of these verbs were modified by time adverbials with known granularity. null The algorithm made the tbllowing choices tbr these data. The second column shows the total number of tensed verbs, the third column only counts those modified by time adverbials with granularity.</Paragraph>
    <Paragraph position="2">  The described algorithm only inspects linguistic factors. Doxnain-specific information could potentially improve results. In the Verbmobil domain e.g. several event types do not hal)i)en in  the present but only in the fllture (tra.vc\], meet, eat, ...).</Paragraph>
    <Paragraph position="3"> 1,1 93.6% of the cases a specific wide-scope time adverbial could be determined with gramflarity constraints. In 4.7% of the cases several time adverbials of equal granularity had wide scope. A good deal of these cases were alignment errors with the translations (e.g. iibel&amp;quot;morgen the day after tomorrow). Other cases were due to the lack of a treatment for coordination (e.g. on Monday aud on Thmwday or ti'om June to August). Some cases were genuine double descriptions of days: (31) a. Is it possible for you tomorrow on the second? b. I would have time on Wednesdw on Wednesday the third of May.</Paragraph>
    <Paragraph position="4"> In 1.7% of the eases the wide-scope adverbial could not be determined because some a.dverbials had overlapping gralmlarity values. Here the main culprit was the unspecified adverb when (see (32a)). Other cases were due to incorrect preposition attachment (see (321,)). (32) a. When shall we meet on Mond~y? Next week/6:30.</Paragraph>
    <Paragraph position="5"> b. Would you be available in the time period until June?</Paragraph>
  </Section>
  <Section position="8" start_page="717" end_page="717" type="evalu">
    <SectionTitle>
9 Summary
</SectionTitle>
    <Paragraph position="0"> The pal/er has presented a disambiguation algorithm ibr translation of German present into English. After a discussion of the factors involved, particular emphasis was placed on an account of scope resolution among time adverbials. It has been shown that grmmhu'ity calculations go a long way towards the goal of full scope resolutioii. The cross-commotions between granularity and scope have been analysed in detail, and some motivation for these connections has been given.</Paragraph>
    <Paragraph position="1"> One area of future work is to apply the model to larger corpora and extend it to cover the full set of tenses. If translations can be aligned with the training data, it would be interesting to investigate the extent to which the model can be used to extract (parts of) the pertinent granularity information on temporal nouns from the corpus (Schiehlen, 1998). For example, the occurrence of a configuration like (33) could be interpreted as evidence tbr NOUN having coarser granularity than week.</Paragraph>
  </Section>
class="xml-element"></Paper>