<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-3009">
  <Title>A Hybrid Approach to Biomedical Named Entity Recognition and Semantic Role Labeling</Title>
  <Section position="5" start_page="245" end_page="245" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> NER and SRL are two key topics in biomedical NLP. For NER, we find broad linguistic features and integrate them into our CRF framework. Our system outperforms most machine learning-based systems, especially in the recognition of protein names (78.4% of F-score). In the future, templates that can match long contextual relations and coordinated NEs may be applied to NER postprocessing. Web corpora may also be used to enhance unknown NE detection. In Bio-SRL, our contribution is threefold. First, we construct a biomedical proposition bank, BioProp, on top of the popular biomedical GENIA treebank following the PropBank annotation scheme. We employ semi-automatic annotation using an SRL system trained on PropBank thereby significantly reducing annotation effort. Second, we construct SEROW, which uses BioProp as its training corpus. Thirdly, we develop a method to automatically generate templates that can boost overall performance, especially on location, manner, adverb, and temporal arguments. In the future, we will expand BioProp to include more biomedical verbs and will also integrate a parser into SEROW.</Paragraph>
  </Section>
class="xml-element"></Paper>