<?xml version="1.0" standalone="yes"?>
<Paper uid="J79-1004">
  <Title>American Journal of Computational Linguistics</Title>
  <Section position="13" start_page="10" end_page="10" type="ackno">
    <SectionTitle>
APPENDIX I1
</SectionTitle>
    <Paragraph position="0"> Some Ideas for the Program to Investigate the Relationship Coverinq Set Size versus ~lax~mum Definition Lenqth The second proposed problem, viz. finding vR as. a function of N for fixed values of vS, is diccussed now. This will he a  task of proportions no less than the present, except for construction of the data base. The following ~rocedure, represented hy a simplified flow chart in Figure 4, is suggesteci for carrying out this task.</Paragraph>
    <Paragraph position="1"> INSERT FIGURE 4 ABOUT HER3 The program starts with known values of Y and v~ (in this  case 22 and 1,464,respectively). It first replaces words in F  having a definition length of 1 (except, of course, those defined by themselves) by their definition in all definitions. Then the program looks for words of short detinition length in R (x = m 2,3,4, etc.) . It substitutes their definition for them in all definitions and counts them out from v .</Paragraph>
    <Paragraph position="2">  records the fmlue. The process is repeated with reclucecl dictionaries, which have different vs values.</Paragraph>
    <Paragraph position="3">  As pointed out earlier it is not suggested that definitions so created are usable or acceptable to the speaker of a natural language. The procedure, however, will produce the numerical relationships des,ired.</Paragraph>
    <Paragraph position="4"> The existing data base, together with its reduce&amp; versions, has been stored on magnetic tape. and is ready to be used as input into the propos~d procedure.</Paragraph>
  </Section>
class="xml-element"></Paper>