<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-1053">
  <Title>A Word-Order Database for Testing Computational Models of Language Acquisition</Title>
  <Section position="5" start_page="8" end_page="8" type="concl">
    <SectionTitle>
4 Conclusion and future work
</SectionTitle>
    <Paragraph position="0"> The thrust of our current research is directed at collecting data for a comprehensive, comparative study of psycho-computational models of syntax acquisition. To support this endeavor, we have developed the Language Domain Database - a publicly available test-bed for studying acquisition models from diverse paradigms.</Paragraph>
    <Paragraph position="1"> Mathematical analysis has shown that learners are extremely sensitive to various distributions in the input stream (Niyogi &amp; Berwick, 1996; Sakas, 2000, 2003). Approaches that thrive in one domain may dramatically flounder in others. So, whether a particular computational model is successful as a model of natural language acquisition is ultimately an empirical issue and depends on the exact conditions under which the model performs well and the extent to which those favorable conditions are in line with the facts of human language. The LDD is a useful tool that can be used within such an empirical research program.</Paragraph>
    <Paragraph position="2"> Future work: Though the LDD has been validated against CHILDES data in certain respects, we intend to extend this work by adding distributions to the LDD that correspond to actual distributions of child-directed speech. For example, what percentage of utterances, in child-directed Japanese, contain pro-drop? object-drop? How often in English does the pattern: S[+WH] aux Verb O1 occur and at what periods of a child's development? We believe that these distributions will shed some light on many of the complex subtleties involved in ambiguity disambiguation and the role of nondeterminism and statistics in the language acquisition process. This is proving to be a formidable, yet surmountable task; one that we are just beginning to tackle.</Paragraph>
  </Section>
class="xml-element"></Paper>