<?xml version="1.0" standalone="yes"?>
<Paper uid="P99-1058">
  <Title>A semantically-derived subset of English for hardware verification</Title>
  <Section position="4" start_page="451" end_page="452" type="intro">
    <SectionTitle>
3 Data
</SectionTitle>
    <Paragraph position="0"> One of our key tasks has been to collect an initial sample of specifications in English, so as to identify linguistic constructions and usages typical of specification discourse. We currently have a corpus of around a hundred sentences, most of which were elicited by asking suitably qualified respondents to describe the behaviour manifested by timing diagrams. An example of such a diagram is displayed in Figure 3, which is adapted from one of Fisler's (1996, p. 5).</Paragraph>
    <Paragraph position="1"> The horizontal axis of the diagram indicates the passing of time (as measured by clock cycles) and the vertical axis indicates the transition of signals between the states of high and low. (A signal is</Paragraph>
    <Paragraph position="3"> for all paths, at the next state c is true for all paths, globally b is true for all paths, eventually there is a state from which, for all paths, at the following state a and b are true</Paragraph>
    <Paragraph position="5"/>
    <Paragraph position="7"> a time-varying value present at some point in the circuit.) In Figure 3, the input signal i makes a transition from high to low which after a one-cycle delay triggers a unit-duration pulse on the output signal o.</Paragraph>
    <Paragraph position="8"> (la-b) give two possible English descriptions of the regularity illustrated by Figure 3,  (1) a. A pulse of width one is generated on the output o one cycle after it detects a falling edge on input i.</Paragraph>
    <Paragraph position="9"> b. If i is high and then is low on the next  cycle, then o is low and after one cycle becomes high and then after one more cycle becomes low.</Paragraph>
    <Paragraph position="10"> while (2) is a CTL description.</Paragraph>
    <Paragraph position="11"> (2) AG(i --+ AX(&amp;quot;,i --+ (--,oAAX(oAAX-,o)))) A noteworthy difference between the two English renderings is that the first is clearly more abstract than the second. Description (lb) is closer to the CTL formula (2), and consequently easier to translate into CTL. 4 For another example of the same phenomenon, consider the timing diagram in Figure 4. As before, sentences (3a-b) give two possible English descriptions of the regularity illustrated by Figure 4, 4Our system does not yet resolve anaphoric references, as in (la). There are existing English-to-CTL systems which do, however, such as that of Nelken and Francez (1996).  (3) a. Every request is eventually acknowledged and once a request is acknowledged the request is eventually deasserted and eventually after that the acknowledge signal goes low.</Paragraph>
    <Paragraph position="12"> b. If r rises then after one cycle eventually a  rises and then after one cycle eventually r falls and then after one cycle eventually a falls.</Paragraph>
    <Paragraph position="13"> which can be rendered in CTL as (4).</Paragraph>
    <Paragraph position="15"> CTL than its (a) counterpart. Nevertheless, (3b) is ontologically richer than CTL in an important respect, in that it makes reference to the event predicates rise and fall.</Paragraph>
  </Section>
class="xml-element"></Paper>