<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1528">
  <Title>k-NN for Local Probability Estimation in Generative Parsing Models</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> This paper describes a generative probabilistic model for parsing, based on Collins (1999), which re-estimates the probability of each parse generated by an initial base parser (Bikel, 2004) using memory-based techniques to estimate local probabilities.</Paragraph>
    <Paragraph position="1"> We used Bikel's re-implementation of the Collins parser (Bikel, 2004) to produce the n-best parses of sentences from the Penn treebank. We then recalculated the probability of each parse tree using a probabilistic model very similar to Collins (1999) Model 1. In addition to the local estimation technique used, our model differs from Collins (1999) Model 1 in that we extend the feature sets used to predict parse structure to include more features from the parse history, and we further decompose some of the model's parameter classes.</Paragraph>
  </Section>
class="xml-element"></Paper>