<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1308">
  <Title>IntEx: A Syntactic Role Driven Protein-Protein Interaction Extractor for Bio-Medical Text</Title>
  <Section position="3" start_page="54" end_page="55" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> Information extraction is the extraction of salient facts about pre-specified types of events, entities (Bunescu, Ge et al. 2003) or relationships from free text. Information extraction from free-text utilizes shallow-parsing techniques (Daelemans, Buchholz et al. 1999), Parts-of-Speech tagging(Brill 1992), noun and verb phrase chunking (Mikheev and Finch 1997), verb subject and object relationships (Daelemans, Buchholz et al. 1999), and learned (Califf and Mooney 1998; Craven and Kumlein 1999; Seymore, McCallum et al. 1999) or hand-build patterns to automate the creation of specialized databases.</Paragraph>
    <Paragraph position="1"> Manual pattern engineering approaches employ shallow parsing with patterns to extract the interactions. In the (Ono, Hishigaki et al. 2001) system,  sentences are first tagged using a dictionary based protein name identifier and then processed by a module which extracts interactions directly from complex and compound sentences using regular expressions based on part of speech tags.</Paragraph>
    <Paragraph position="2"> The SUISEKI system of Blaschke (Blaschke, Andrade et al. 1999) also uses regular expressions, with probabilities that reflect the experimental accuracy of each pattern to extract interactions into predefined frame structures.</Paragraph>
    <Paragraph position="3"> GENIES (Friedman, Kra et al. 2001) utilizes a grammar based NLP engine for information extraction. Recently, it has been extended as GeneWays (Rzhetsky, Iossifov et al. 2004), which also provides a Web interface that allows users to search and submit papers of interest for analysis. The BioRAT system (Corney, Buxton et al. 2004) uses manually engineered templates that combine lexical and semantic information to identify protein interactions. The GeneScene system(Leroy, Chen et al. 2003) extracts interactions using frequent preposition-based templates.</Paragraph>
    <Paragraph position="4"> Grammar engineering approaches, on the other hand use manually generated specialized grammar rules (Rinaldi, Schneider et al. 2004) that perform a deep parse of the sentences. Temkin (Temkin and Gilder 2003) addresses the problem of extracting protein interactions by using an extendable but manually built Context Free Grammar (CFG) that is designed specifically for parsing biological text. The PathwayAssist system uses an NLP system, MedScan (Novichkova, Egorov et al. 2003), for the biomedical domain that tags the entities in text and produces a semantic tree. Slot filler type rules are engineered based on the semantic tree representation to extract relationships from text. Recently, extraction systems have also used link grammar (Grinberg, Lafferty et al. 1995) to identify interactions between proteins (Ding, Berleant et al. 2003). Their approach relies on various linkage paths between named entities such as gene and protein names. Such manual pattern engineering approaches for information extraction are very hard to scale up to large document collections since they require labor-intensive and skill-dependent pattern engineering.</Paragraph>
    <Paragraph position="5"> Machine learning approaches have also been used to learn extraction rules from user tagged training data. These approaches represent the rules learnt in various formats such as decision trees (Chiang, Yu et al. 2004) or grammar rules (Phuong, Lee et al.</Paragraph>
    <Paragraph position="6"> 2003). Craven et al (Craven and Kumlien 1999) explored an automatic rule-learning approach that uses a combination of FOIL (Quinlan 1990) and Naive Bayes Classifier to learn extraction rules.</Paragraph>
  </Section>
class="xml-element"></Paper>