<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3312">
  <Title>Postnominal Prepositional Phrase Attachment in Proteomics</Title>
  <Section position="9" start_page="88" end_page="88" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> The next step for BioNLP is to process the full text of scienti c articles, where heavy NPs with potentially long chains of PP attachments are frequent.</Paragraph>
    <Paragraph position="1"> This study has investigated the attachment behavior of postnominal PPs in enzyme-related texts and evaluated a small set of simple attachment heuristics on a test set of over 3000 PPs from a collection of more varied texts in proteomics. The heuristics cover all prepositions, even infrequent ones, that nonetheless convey important information. This approach requires only NP chunked input and a nominalization dictionary, all readily available from on-line resources. The heuristics are thus useful for shallow approaches and their accuracy of 82% puts them in a position to reliably improve both, proper recognition of entities and their properties and bottom-up recognition of relationships between entities expressed in nominalizations.</Paragraph>
  </Section>
class="xml-element"></Paper>