template

nllb.py

@@ -124,8 +124,10 @@ class NLLB(datasets.GeneratorBasedBuilder):
                 "source_sentence_lid": datasets.Value("float32"),
                 "target_sentence_lid": datasets.Value("float32"),
                 "source_sentence_source": datasets.Value("string"),
+                "source_sentence_perplexity": datasets.Value("float32"),
                 "source_sentence_url": datasets.Value("string"),
                 "target_sentence_source": datasets.Value("string"),
+                "target_sentence_perplexity": datasets.Value("float32"),
                 "target_sentence_url": datasets.Value("string"),
             }
         )
@@ -167,14 +169,14 @@ class NLLB(datasets.GeneratorBasedBuilder):
             )
         ]
 
-    def _generate_examples(self, filepath, source_lg, target_lg):
+    def _generate_examples(self, filepath, source_lg, target_lg, source_max_perplexity=None, target_max_perplexity=None):
         if self.config.source == "statmt":
             # MT stats didn't published all the metadata
             return self._generate_minimal_examples(filepath, source_lg, target_lg)
 
-        return self._generate_full_examples(filepath, source_lg, target_lg)
+        return self._generate_full_examples(filepath, source_lg, target_lg, source_max_perplexity=source_max_perplexity, target_max_perplexity=target_max_perplexity)
 
-    def _generate_full_examples(self, filepath, source_lg, target_lg):
+    def _generate_full_examples(self, filepath, source_lg, target_lg, source_max_perplexity=None, target_max_perplexity=None):
         with open(filepath, encoding="utf-8") as f:
             # reader = csv.reader(f, delimiter="\t")
             for id_, example in enumerate(f):
@@ -186,6 +188,9 @@ class NLLB(datasets.GeneratorBasedBuilder):
                         source_lg: datarow[0],
                         target_lg: datarow[1],
                     }
+
+                    row["source_sentence_perplexity"] = None # TODO: compute the perplexity for the source sentence here
+                    row["target_sentence_perplexity"] = None # TODO: compute the perplexity for the target sentence here
                     row["laser_score"] = float(datarow[2])
                     row["source_sentence_lid"] = float(datarow[3])
                     row["target_sentence_lid"] = float(datarow[4])
@@ -198,6 +203,10 @@ class NLLB(datasets.GeneratorBasedBuilder):
                 except:
                     print(datarow)
                     raise
+                if source_max_perplexity is not None and row["source_sentence_perplexity"] > source_max_perplexity:
+                    continue
+                if target_max_perplexity is not None and row["target_sentence_perplexity"] > target_max_perplexity:
+                    continue
                 yield id_, row
 
     def _generate_minimal_examples(self, filepath, source_lg, target_lg):