<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2039">
  <Title>Parsing Aligned Parallel Corpus by Projecting Syntactic Relations from Annotated Source Corpus</Title>
  <Section position="7" start_page="306" end_page="307" type="evalu">
    <SectionTitle>
6 Experimental Results
</SectionTitle>
    <Paragraph position="0"> Currently the system can handle the following types of phrases in different simple sentences.</Paragraph>
    <Paragraph position="1"> Noun Phrase. There can be four basic elements of an English NP6: determiner, pre-modifier, noun (essential), post-modifier. The system can handle any combination of the following: adjective, noun, present participle or past participle as premodifier, and adjective, present participle, past participle or preposition phrase as post-modifier.</Paragraph>
    <Paragraph position="2"> Note that some of these cases may be translated as complex sentence in Hindi (e.g., (book on the table [?] jo kitaab mej par rakhii hai). We are working upon such cases.</Paragraph>
    <Paragraph position="3">  Verb Phrase. The system can handle all the four aspects (indefinite, continuous, perfect and perfect continuous) for all three tenses. Other cases of VPs (e.g., modals, passives, compound verbs) can be handled easily by just identifying and putting the corresponding auxiliary verbs and their linking requirements in the auxiliary verb list.</Paragraph>
    <Paragraph position="4"> Since the system is not fully automated yet, we could not test our system on a large corpus. The system has been tested on about 200 sentences following the specific phrase structures mentioned above. These sentences have been taken randomly from translation books, stories books and advertisement materials. These sentences were manually parsed and a total of 1347 links were obtained. These links were compared with the system's output. Table 3 summarizes the findings.</Paragraph>
    <Paragraph position="5">  After analyzing the results, we found that * For some links, suffixes were wrong. This was due to insufficiency of rules identifying morphological information.</Paragraph>
    <Paragraph position="6"> * Due to incompleteness of some cases of ParseFromSpecial() module, some wrong links were assigned. Also, some links which should not have been projected, were projected in the Hindi sentence. We are working towards exploring these cases in detail.</Paragraph>
    <Paragraph position="7"> * Some links were found missing in the parsing since corresponding sentence structures are yet to be considered in the scheme.</Paragraph>
  </Section>
class="xml-element"></Paper>