Merge pull request #98 from huridocs/labeled-as-empty

Support labeled data with no value

requirements.txt

@@ -1,5 +1,5 @@
-git+https://github.com/huridocs/pdf-document-layout-analysis@a834f712edf0fc0a4660de0270f3efbb08e0031c
-git+https://github.com/huridocs/queue-processor@1875372bf9f6dcd1995a32c4e50ff92aa45f9ea8
+git+https://github.com/huridocs/pdf-document-layout-analysis@6262a95d1dd09055556a8e081eacfe02dd308b89
+git+https://github.com/huridocs/queue-processor@716ddf050c59035583b0852dc0b78a7860ce5c05
 slugify==0.0.1
 python-Levenshtein==0.25.1
 tdda==2.0.9
@@ -15,7 +15,7 @@ mongomock==4.1.2
 fasttext-wheel==0.9.2
 rich==13.7.1
 joblib==1.4.0
-tqdm==4.66.2
+tqdm==4.66.3
 rapidfuzz==3.8.1
 sentry_sdk==1.44.0
 pymongo==4.6.3

src/Extractor.py

@@ -186,9 +186,11 @@ def get_suggestions(self) -> list[Suggestion]:
             if extractor_instance.get_name() != extractor_name:
                 continue
 
-            message = f"Using {extractor_instance.get_name()} to calculate {len(prediction_samples)} suggestions"
+            suggestions = extractor_instance.get_suggestions(prediction_samples)
+            suggestions = [suggestion.mark_suggestion_if_empty() for suggestion in suggestions]
+            message = f"Using {extractor_instance.get_name()} to calculate {len(suggestions)} suggestions"
             send_logs(self.extraction_identifier, message)
-            return extractor_instance.get_suggestions(prediction_samples)
+            return suggestions
 
         send_logs(self.extraction_identifier, f"No extractor available", Severity.error)
         return []

src/data/LabeledData.py

@@ -11,6 +11,7 @@ class LabeledData(BaseModel):
     entity_name: str = ""
     language_iso: str = ""
     label_text: str = ""
+    empty_value: bool = False
     values: list[Option] = list()
     source_text: str = ""
     page_width: float = 0

src/data/Suggestion.py

@@ -17,11 +17,24 @@ class Suggestion(BaseModel):
     xml_file_name: str = ""
     entity_name: str = ""
     text: str = ""
+    empty_suggestion: bool = False
     values: list[Option] = list()
     segment_text: str = ""
     page_number: int = 1
     segments_boxes: list[SegmentBox] = list()
 
+    def is_empty(self):
+        if self.empty_suggestion:
+            return True
+
+        return not self.text and not self.values
+
+    def mark_suggestion_if_empty(self):
+        if self.is_empty():
+            self.empty_suggestion = True
+
+        return self
+
     def to_dict(self):
         suggestion_dict = self.model_dump()
         suggestion_dict["segments_boxes"] = [x.to_dict() for x in self.segments_boxes]

src/extractors/ToTextExtractor.py

@@ -10,13 +10,6 @@
 from data.Suggestion import Suggestion
 from extractors.ExtractorBase import ExtractorBase
 from extractors.ToTextExtractorMethod import ToTextExtractorMethod
-from extractors.text_to_text_extractor.methods.DateParserMethod import DateParserMethod
-from extractors.text_to_text_extractor.methods.DateParserWithBreaksMethod import DateParserWithBreaksMethod
-from extractors.text_to_text_extractor.methods.InputWithoutSpaces import InputWithoutSpaces
-from extractors.text_to_text_extractor.methods.MT5TrueCaseEnglishSpanishMethod import MT5TrueCaseEnglishSpanishMethod
-from extractors.text_to_text_extractor.methods.RegexMethod import RegexMethod
-from extractors.text_to_text_extractor.methods.RegexSubtractionMethod import RegexSubtractionMethod
-from extractors.text_to_text_extractor.methods.SameInputOutputMethod import SameInputOutputMethod
 from send_logs import send_logs
 
 RETRAIN_SAMPLES_THRESHOLD = 250
@@ -115,8 +108,8 @@ def get_best_method(self, extraction_data: ExtractionData):
             try:
                 performance = method_instance.performance(training_set, test_set)
             except Exception as e:
-                message = f"Error checking {method_instance.get_name()}: {e}"
-                send_logs(self.extraction_identifier, message, Severity.error)
+                message = f"Error checking {method_instance.get_name()}"
+                send_logs(self.extraction_identifier, message, Severity.error, e)
                 performance = 0
             performance_log += f"{method_instance.get_name()}: {round(performance, 2)}%\n"
             send_logs(self.extraction_identifier, f"Performance {method_instance.get_name()}: {performance}%")

src/extractors/pdf_to_multi_option_extractor/PdfToMultiOptionExtractor.py

@@ -211,7 +211,7 @@ def get_method_performance(
             performance = method.get_performance(train_set, test_set)
         except Exception as e:
             severity = Severity.error if method.REPORT_ERRORS else Severity.info
-            send_logs(self.extraction_identifier, f"Error checking {method.get_name()}: {e}", severity)
+            send_logs(self.extraction_identifier, f"Error checking {method.get_name()}", severity, e)
             performance = 0
 
         self.reset_extraction_data(train_set)