<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-0603">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics How and Where do People Fail with Time: Temporal Reference Mapping Annotation by Chinese and English Bilinguals Yang Ye SS</Title>
  <Section position="3" start_page="0" end_page="13" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> In recent years, the research community has seen a fast-growing volume of work in temporal information processing. Consequently, the investigation and practice of temporal information annotation by human experts have emerged from the corpus annotation research. To evaluate automatic temporal relation classification systems, annotated corpora must be created and validated, which motivates experiments and research in temporal information annotation.</Paragraph>
    <Paragraph position="1"> One important temporal relation distinction that human beings make is the temporal reference distinction based on relative positioning between the following three time parameters, as proposed by (Reichenbach, 1947): speech time (S), event time (E) and reference time (R). Temporal reference distinction is linguistically realized as tenses. Languages have various granularities of tense representations; some have finer-grained tenses or aspects than others. This poses a great challenge to automatic cross-lingual tense mapping. The same challenge holds for cross-lingual tense annotation, especially for language pairs that have dramatically different tense strategies. A decent solution for cross-lingual tense mapping will benefit a variety of NLP tasks such as Machine Translation, Cross-lingual Question Answering (CLQA), and Multi-lingual Information Summarization. While automatic cross-lingual tense mapping has recently started to receive research attention, such as in (Olsen,et al., 2001) and (Ye, et al., 2005), to the best of our knowledge, human performance on tense and aspect annotation for machine translation between English and Chinese has not received any systematic investigation to date. Cross-linguistic NLP tasks, especially those requiring a more accurate tense and aspect resolution, await a more focused study of human tense and aspect annotation performance.</Paragraph>
    <Paragraph position="2"> Chinese and English are a language pair in which tense and aspect are represented at different levels of units: one being realized at the word level and the other at the morpheme level.</Paragraph>
    <Paragraph position="3"> This paper reports on a series of cross-linguistic tense annotation experiments between Chinese and English, and provides statistical inference for different linguistic factors via a series of statistical modeling. Since tense and aspect are morphologically merged in English, tense annotation  discussed in this paper also includes elements of aspect. We only deal with tense annotation in Chinese-to-English scenario in the scope of this paper.</Paragraph>
    <Paragraph position="4"> The remaining part of the paper is organized as follows: Section 2 summarizes the significant related works in temporal information annotation and points out how this study relates to yet differs from them. Section 3 reports the details of three tense annotation experiments under three scenarios. Section 4 discusses the inter-judge agreement by presenting two measures of agreement: the Kappa Statistic and accuracy-based measurement. Section 5 investigates and reports on the significance of different linguistic factors in tense annotation via an ANOVA analysis, a logistic regression analysis and a log-linear model analysis.</Paragraph>
    <Paragraph position="5"> Finally, section 6 concludes the paper and points out directions for future research.</Paragraph>
  </Section>
class="xml-element"></Paper>