<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2017">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Analysis and Synthesis of the Distribution of Consonants over Languages: A Complex Network Approach</Title>
  <Section position="8" start_page="133" end_page="134" type="concl">
    <SectionTitle>
6 Conclusions, Discussion and Future
</SectionTitle>
    <Paragraph position="0"> Work In this paper, we have analyzed and synthesized the consonant inventories of the world's languages in terms of a complex network. We dedicated the preceding sections essentially to, * Represent the consonant inventories through a bipartite network called PlaNet, * Provide a systematic study of certain important properties of the consonant inventories with the help of PlaNet, * Propose analytical explanations for the two regime power law curves (obtained from PlaNet) on the basis of the distribution of the consonant inventory size over languages together with the principle of preferential attachment, null 3Mean error is defined as the average difference between the ordinate pairs where the abscissas are equal. * Provide a simplified mathematical model to support our analytical explanations, and * Develop a synthesis model for PlaNet based on preferential attachment where the consonant inventory size distribution is known a priori.</Paragraph>
    <Paragraph position="1"> We believe that the general explanation provided here for the two regime power law is a fundamental result, and can have a far reaching impact, because two regime behavior is observed in many other networked systems.</Paragraph>
    <Paragraph position="2"> Until now we have been mainly dealing with the computational aspects of the distribution of consonants over the languages rather than exploring the real world dynamics that gives rise to such a distribution. An issue that draws immediate attention is that how preferential attachment, which is a general phenomenon associated with network evolution, can play a prime role in shaping the consonant inventories of the world's languages. The answer perhaps is hidden in the fact that language is an evolving system and its present structure is determined by its past evolutionary history. Indeed an explanation based on this evolutionary model, with an initial disparity in the distribution of consonants over languages, can be intuitively verified as follows - let there be a language community of N speakers communicating among themselves by means of only two consonants say /k/ and /g/.</Paragraph>
    <Paragraph position="3"> If we assume that every speaker has l descendants and language inventories are transmitted with high fidelity, then after i generations it is expected that the community will consist of mli /k/ speakers and nli /g/ speakers. Now if m &gt; n and l &gt; 1, then for sufficiently large i, mli greatermuch nli. Stated differently, the /k/ speakers by far outnumbers the /g/ speakers even if initially the number of /k/ speakers is only slightly higher than that of the /g/ speakers. This phenomenon is similar to that of preferential attachment where language communities get attached to, i.e., select, consonants that are already highly preferred. Nevertheless, it remains to be seen where from such an initial disparity in the distribution of the consonants over languages might have originated.</Paragraph>
    <Paragraph position="4"> In this paper, we mainly dealt with the occurrence principles of the consonants in the inventories of the world's languages. The work can be further extended to identify the co-occurrence likelihood of the consonants in the language inventories  and subsequently identify the groups or communities within them. Information about such communities can then help in providing an improved insight about the organizing principles of the consonant inventories.</Paragraph>
  </Section>
class="xml-element"></Paper>