summaryrefslogtreecommitdiff
path: root/extractor/sampler.cc
diff options
context:
space:
mode:
authorPaul Baltescu <pauldb89@gmail.com>2013-11-26 15:01:14 +0000
committerPaul Baltescu <pauldb89@gmail.com>2013-11-26 15:01:14 +0000
commita3826db61847a55f59bb9666f61fd1bb88888085 (patch)
tree022475bafbf71ba6aaeb98efdbafcde24f7e60a5 /extractor/sampler.cc
parent1cd86c44e1799c441cdcda2a022be0ee6e52d38c (diff)
Script for grammar extraction only.
Diffstat (limited to 'extractor/sampler.cc')
-rw-r--r--extractor/sampler.cc23
1 files changed, 11 insertions, 12 deletions
diff --git a/extractor/sampler.cc b/extractor/sampler.cc
index fc386ed1..887aaec1 100644
--- a/extractor/sampler.cc
+++ b/extractor/sampler.cc
@@ -15,6 +15,7 @@ Sampler::~Sampler() {}
PhraseLocation Sampler::Sample(
const PhraseLocation& location,
const unordered_set<int>& blacklisted_sentence_ids) const {
+ shared_ptr<DataArray> source_data_array = suffix_array->GetData();
vector<int> sample;
int num_subpatterns;
if (location.matchings == NULL) {
@@ -22,32 +23,30 @@ PhraseLocation Sampler::Sample(
num_subpatterns = 1;
int low = location.sa_low, high = location.sa_high;
double step = max(1.0, (double) (high - low) / max_samples);
- double i = low, last = i;
- bool found;
- shared_ptr<DataArray> source_data_array = suffix_array->GetData();
+ double i = low, last = i - 1;
while (sample.size() < max_samples && i < high) {
int x = suffix_array->GetSuffix(Round(i));
int id = source_data_array->GetSentenceId(x);
+ bool found = false;
if (blacklisted_sentence_ids.count(id)) {
- found = false;
- double backoff_step = 1;
- while (true) {
- if ((double)backoff_step >= step) break;
+ for (int backoff_step = 1; backoff_step <= step; ++backoff_step) {
double j = i - backoff_step;
x = suffix_array->GetSuffix(Round(j));
id = source_data_array->GetSentenceId(x);
if (x >= 0 && j > last && !blacklisted_sentence_ids.count(id)) {
- found = true; last = i; break;
+ found = true;
+ last = i;
+ break;
}
double k = i + backoff_step;
x = suffix_array->GetSuffix(Round(k));
id = source_data_array->GetSentenceId(x);
- if (k < min(i+step, (double)high) &&
+ if (k < min(i+step, (double) high) &&
!blacklisted_sentence_ids.count(id)) {
- found = true; last = k; break;
+ found = true;
+ last = k;
+ break;
}
- if (j <= last && k >= high) break;
- backoff_step++;
}
} else {
found = true;