<?xml version="1.0" standalone="yes"?>
<Paper uid="P98-1044">
  <Title>Veins Theory: A Model of Global Discourse Cohesion and Coherence</Title>
  <Section position="3" start_page="282" end_page="282" type="intro">
    <SectionTitle>
2 Accessibility
</SectionTitle>
    <Paragraph position="0"> The domain of accessibility of a unit is defined as the string of units appearing in its vein expression and prefixing that unit itself.</Paragraph>
    <Paragraph position="1"> More formally, for each terminal node u, if vein(u) is its vein, then accessibility from u is given by acc(u) = pref(u, unmark(vein(u)), where: vein is the function that computes the vein; unmark(x) is a function that removes the markers from all symbols of its argument; * pref is a function that retains the prefix of the second argument up to and including the first argument (e.g., if a and 13 are strings of labels and u is a label, pref(u, aufl) = ocu, Conjecture CI: References from a given unit are possible only in its domain of accessibility. In particular, we can say the following:  1. In most cases, if B is a unit and beB is a referential expression, then either b directly realizes a center that appears for the first time in the discourse, or it refers back to another center realized by a referential expression aeA, such that Aeacc(B). 2 Such cases instantiate direct references.</Paragraph>
    <Paragraph position="2"> 2. If (1) is not applicable, then if A, B, and C are units, ceC is a referential expression that refers to beB, and B is not on the vein of C .(i.e., it is not visible from C), then there is an item aeA, where A is a unit on the common vein of B and C, such that both b and c refer to a. In this case we say that c is an indirect reference to aJ 3. If neither (1) nor (2) is applicable, then the  reference in C can be understood without the referee, as if the corresponding entity were introduced in the discourse for the first time. Such references are inferential references.</Paragraph>
    <Paragraph position="3"> Note that VT is applicable even when the division into units ~s coarser than in our examples. For instance, Example 1 in its entirety could be taken to comprise a single unit; if it appeared in the context of a larger discourse, it would still be possible to.compute its veins (although, of course, the veins would likely be shorter because there are fewer units to consider). It can be proven formally (Cristea,  2 If a and b are referential expressions, where the center (directly) realized by b is the same as the one (directly) realized by a, or where it is a role of the center (directly) realized by a, we will say that b refers (back) to a, or b is a bridge reference to a. 3 On the basis of their common semantic  1998) that when passing from a finer granularity to a coarser one the accessibility constraints are still obeyed. This observation is important in relation to other approaches that search for stability with respect to granularity (see for instance, Walker, 1996).</Paragraph>
  </Section>
class="xml-element"></Paper>