No public description

PiperOrigin-RevId: 674062722

nisaba/translit/tools/calculate_error_rate.cc

@@ -487,6 +487,9 @@ void MultiRefErrorRate::CalculateErrorRate(bool pairwise_edits) {
 void MultiRefErrorRate::CalculateErrorRate(absl::string_view reffile,
                                            absl::string_view testfile,
                                            bool pairwise_edits) {
+  if (output_syms_.Find("<epsilon>") < 0) {
+    output_syms_.AddSymbol("<epsilon>");
+  }
   ReadInputs(reffile, /*is_reference=*/true);
   ReadInputs(testfile, /*is_reference=*/false);
   CalculateErrorRate(pairwise_edits);

nisaba/translit/tools/calculate_error_rate.h

@@ -49,7 +49,7 @@ struct EditDistanceDouble {
   double insertions = 0.;
   double deletions = 0.;
 
-  EditDistanceDouble() {}
+  EditDistanceDouble() = default;
 
   explicit EditDistanceDouble(const EditDistanceInt &x)
       : reference_length(x.reference_length),
@@ -101,25 +101,27 @@ struct EditDistanceDouble {
 // Class for calculating error rates possibly from multiple references.
 class MultiRefErrorRate {
  public:
-  MultiRefErrorRate() = default;
-
-  explicit MultiRefErrorRate(bool is_split_chars)
-      : is_split_chars_(is_split_chars) {
-    output_syms_.AddSymbol("<epsilon>");
-  }
+  explicit MultiRefErrorRate(bool is_split_chars = true)
+      : is_split_chars_(is_split_chars) {}
 
   // Calculates multi reference error rate for given test file. If pairwise_edit
   // is set to true, provides pairwise edits between all reference and test
   // outputs for the same input.
   void CalculateErrorRate(absl::string_view reffile, absl::string_view testfile,
-                          bool pairwise_edits);
+                          bool pairwise_edits = false);
 
   // Writes results to output file.
   void Write(absl::string_view ofile, bool pairwise_edits);
 
   // Calculates error rate and optionally writes to provide file pointer.
   double CalcErrorRate();
 
+ protected:
+  // Returns tokenized string associated with idx and k-th item from test_input_
+  // if is_test_item is true, otherwise references_.
+  std::vector<std::string> GetTokenizedString(int idx, int k,
+                                              bool is_test_item = false) const;
+
  private:
   // Reads in TSV input file, either reference or test file.
   // Column 1 should be the example index, starting from 0.
@@ -128,11 +130,6 @@ class MultiRefErrorRate {
   // For the reference file this is a count; for test a -log prob.
   void ReadInputs(absl::string_view input_file, bool is_reference);
 
-  // Returns tokenized string associated with idx and k-th item from test_input_
-  // if is_test_item is true, otherwise references_.
-  std::vector<std::string> GetTokenizedString(int idx, int k,
-                                              bool is_test_item) const;
-
   // Scans through data set and calculates error rate.
   void CalculateErrorRate(bool pairwise_edits);
 

nisaba/translit/tools/calculate_error_rate_test.cc

@@ -28,6 +28,12 @@ namespace {
 
 constexpr float kFloatDelta = 0.00001;  // Delta for float comparisons.
 
+class MultiRefErrorRateHelper : public MultiRefErrorRate {
+ public:
+  using MultiRefErrorRate::GetTokenizedString;
+  MultiRefErrorRateHelper() = default;
+};
+
 class MultiRefErrorRateTest : public ::testing::Test {
  protected:
   // Creates first file for testing error rate.
@@ -71,7 +77,37 @@ class MultiRefErrorRateTest : public ::testing::Test {
   std::string file_two_;  // File name for second k-best output.
 };
 
-// TODO: create test of GetTokenizedString to test bad indices.
+// To test behavior with out-of-range indices.
+TEST_F(MultiRefErrorRateTest, GetTokenizedStringTests) {
+  // Initializing calculator with file_one_ as reference.
+  MultiRefErrorRateHelper multi_ref_calc;
+  multi_ref_calc.CalculateErrorRate(/*reffile=*/file_one_,
+                                    /*testfile=*/file_two_);
+  // Example 0, item 0 exists in reference, so resulting vector non-empty.
+  EXPECT_FALSE(multi_ref_calc.GetTokenizedString(0, 0).empty());
+  // Example index out of range, so vector empty.
+  EXPECT_TRUE(multi_ref_calc.GetTokenizedString(-1, 0).empty());
+  EXPECT_TRUE(
+      multi_ref_calc.GetTokenizedString(-1, 0, /*is_test_item=*/true).empty());
+  // Example index out of range, so vector empty.
+  EXPECT_TRUE(multi_ref_calc.GetTokenizedString(3, 0).empty());
+  EXPECT_TRUE(
+      multi_ref_calc.GetTokenizedString(3, 0, /*is_test_item=*/true).empty());
+  // Example 1 of reference and test have more than 1 item, vector non-empty.
+  EXPECT_FALSE(multi_ref_calc.GetTokenizedString(1, 1).empty());
+  EXPECT_FALSE(
+      multi_ref_calc.GetTokenizedString(1, 1, /*is_test_item=*/true).empty());
+  // Example 2 of reference and test have just 1 item, hence vector empty.
+  EXPECT_TRUE(multi_ref_calc.GetTokenizedString(2, 1).empty());
+  EXPECT_TRUE(
+      multi_ref_calc.GetTokenizedString(2, 1, /*is_test_item=*/true).empty());
+  // Example 1 of reference has 2 items but test has 3 items, so vector should
+  // be empty for reference but not for test.
+  EXPECT_TRUE(multi_ref_calc.GetTokenizedString(1, 2).empty());
+  EXPECT_FALSE(
+      multi_ref_calc.GetTokenizedString(1, 2, /*is_test_item=*/true).empty());
+}
+
 // Verifies minimum error rate calculation in both directions.
 TEST_F(MultiRefErrorRateTest, CorrectMinErrorRates) {
   // Initializing calculator for use with file_one_ as reference.