summaryrefslogtreecommitdiff
path: root/extractor/precomputation.h
diff options
context:
space:
mode:
authorPaul Baltescu <pauldb89@gmail.com>2013-03-10 01:01:01 +0000
committerPaul Baltescu <pauldb89@gmail.com>2013-03-10 01:01:01 +0000
commit6d43674e6b224281e43ccefc87224a7ba2fbb99a (patch)
tree36e0128b60755e47d217825fca13ccfb3f701158 /extractor/precomputation.h
parent1b9ca189fd0549bd6d969edf618f92ea59184b12 (diff)
Added comments. Hooray!
Diffstat (limited to 'extractor/precomputation.h')
-rw-r--r--extractor/precomputation.h21
1 files changed, 21 insertions, 0 deletions
diff --git a/extractor/precomputation.h b/extractor/precomputation.h
index 2c1eccf8..e3c4d26a 100644
--- a/extractor/precomputation.h
+++ b/extractor/precomputation.h
@@ -20,8 +20,19 @@ typedef unordered_map<vector<int>, vector<int>, VectorHash> Index;
class SuffixArray;
+/**
+ * Data structure wrapping an index with all the occurrences of the most
+ * frequent discontiguous collocations in the source data.
+ *
+ * Let a, b, c be contiguous collocations. The index will contain an entry for
+ * every collocation of the form:
+ * - aXb, where a and b are frequent
+ * - aXbXc, where a and b are super-frequent and c is frequent or
+ * b and c are super-frequent and a is frequent.
+ */
class Precomputation {
public:
+ // Constructs the index using the suffix array.
Precomputation(
shared_ptr<SuffixArray> suffix_array, int num_frequent_patterns,
int num_super_frequent_patterns, int max_rule_span,
@@ -32,6 +43,7 @@ class Precomputation {
void WriteBinary(const fs::path& filepath) const;
+ // Returns a reference to the index.
virtual const Index& GetCollocations() const;
static int FIRST_NONTERMINAL;
@@ -41,14 +53,23 @@ class Precomputation {
Precomputation();
private:
+ // Finds the most frequent contiguous collocations.
vector<vector<int> > FindMostFrequentPatterns(
shared_ptr<SuffixArray> suffix_array, const vector<int>& data,
int num_frequent_patterns, int max_frequent_phrase_len,
int min_frequency);
+
+ // Given the locations of the frequent contiguous collocations in a sentence,
+ // it adds new entries to the index for each discontiguous collocation
+ // matching the criteria specified in the class description.
void AddCollocations(
const vector<std::tuple<int, int, int> >& matchings, const vector<int>& data,
int max_rule_span, int min_gap_size, int max_rule_symbols);
+
+ // Adds an occurrence of a binary collocation.
void AddStartPositions(vector<int>& positions, int pos1, int pos2);
+
+ // Adds an occurrence of a ternary collocation.
void AddStartPositions(vector<int>& positions, int pos1, int pos2, int pos3);
Index collocations;