<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1101">
  <Title>Combining Linguistic Features with Weighted Bayesian Classifier for Temporal Reference Processing</Title>
  <Section position="7" start_page="321" end_page="321" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> Temporal reference processing has received growing attentions in last decades. However this topic has not been well studied in Chinese. In this paper, we proposed a method to determine temporal relations in Chinese by employing linguistic knowledge and machine learning approaches. Thirteen related linguistic features were recognized and temporal indicators were further grouped with respect to grammatical functions or semantic roles. This allows features to be compared with those in the same group. To accommodate the fact that the different types of features support varied importance, we extended Naive Bayesian Classifier to Weighted Bayesian Classifier and applied Simulated Annealing algorithm to optimize weight parameters. To avoid over-fitting problem, K-fold Cross-Validation technique was incorporated to evaluate the objective function of the optimization algorithm. Establishing the temporal relations between two events could be extended to provide a determination of the temporal relations among multiple events in a discourse. With such an extension, this temporal analysis approach could be incorporated into various NLP applications, such as question answering and machine translation.</Paragraph>
  </Section>
class="xml-element"></Paper>