miller/docs6/docs/ngrams/mkv.mlr

83 lines
1.4 KiB
Text
Executable file

begin {
if (is_absent(@olen)) {
@olen = 16;
}
if (is_absent(@ocount)) {
@ocount = 16;
}
}
for (_, v in $*) {
if (string(v) == "inf" || string(v) == "nan") {
continue;
}
int n = strlen(v);
if (n >= 1) {
str a = substr(v, 0, 0);
@a_histo[a] += 1;
for (int i = 0; i < n-1; i += 1) {
str a = substr(v, i, i);
str b = substr(v, i+1, i+1);
@ab_histo[a][b] += 1;
}
}
}
end {
num a_sum = 0;
for (a, na in @a_histo) {
a_sum += na;
}
num a_cumu = 0.0;
for (a, na in @a_histo) {
num a_p = @a_histo[a] / a_sum;
a_cumu += a_p;
@a_pmf[a] = a_p;
@a_cmf[a] += a_cumu;
}
for (a in @ab_histo) {
num ab_sum = 0.0;
for (b, nab in @ab_histo[a]) {
ab_sum += nab;
}
num ab_cumu = 0.0;
for (b, _ in @ab_histo[a]) {
num ab_p = @ab_histo[a][b] / ab_sum;
ab_cumu += ab_p;
@ab_pmf[a][b] = ab_p;
@ab_cmf[a][b] = ab_cumu;
}
}
#dump;
for (int oi = 0; oi < @ocount; oi += 1) {
oa_u = urand();
oa = "?";
for (a, c in @a_cmf) {
if (oa_u < c) {
oa = a;
break;
}
}
str out = oa;
for (int i = 1; i < @olen; i += 1) {
num ab_u = urand();
ob = "???";
for (b, c in @ab_cmf[oa]) {
#print "BC ".b.c;
if (ab_u < c) {
ob = b;
break;
}
}
if (ob == "???") {
break;
}
out .= ob;
oa = ob;
}
print out;
}
}