Merge pull request #3271 from flairNLP/3265-question-sentenceto_dictt…

…ag_type=ner-no-longer-have-the-entities-key

recreate `to_dict` and add relations

flair/data.py

@@ -543,6 +543,14 @@ def set_label(self, typename: str, value: str, score: float = 1.0):
         else:
             DataPoint.set_label(self, typename=typename, value=value, score=score)
 
+    def to_dict(self, tag_type: Optional[str] = None):
+        return {
+            "text": self.text,
+            "start_pos": self.start_position,
+            "end_pos": self.end_position,
+            "labels": [label.to_dict() for label in self.get_labels(tag_type)],
+        }
+
 
 class Span(_PartOfSentence):
     """This class represents one textual span consisting of Tokens."""
@@ -604,6 +612,14 @@ def __len__(self) -> int:
     def embedding(self):
         return self.get_embedding()
 
+    def to_dict(self, tag_type: Optional[str] = None):
+        return {
+            "text": self.text,
+            "start_pos": self.start_position,
+            "end_pos": self.end_position,
+            "labels": [label.to_dict() for label in self.get_labels(tag_type)],
+        }
+
 
 class Relation(_PartOfSentence):
     def __new__(self, first: Span, second: Span):
@@ -664,6 +680,15 @@ def end_position(self) -> int:
     def embedding(self):
         pass
 
+    def to_dict(self, tag_type: Optional[str] = None):
+        return {
+            "from_text": self.first.text,
+            "to_text": self.second.text,
+            "from_idx": self.first.tokens[0].idx - 1,
+            "to_idx": self.second.tokens[0].idx - 1,
+            "labels": [label.to_dict() for label in self.get_labels(tag_type)],
+        }
+
 
 class Sentence(DataPoint):
     """A Sentence is a list of tokens and is used to represent a sentence or text fragment."""
@@ -760,17 +785,17 @@ def __init__(
     def unlabeled_identifier(self):
         return f'Sentence[{len(self)}]: "{self.text}"'
 
-    def get_relations(self, type: str) -> List[Relation]:
+    def get_relations(self, label_type: Optional[str] = None) -> List[Relation]:
         relations: List[Relation] = []
-        for label in self.get_labels(type):
+        for label in self.get_labels(label_type):
             if isinstance(label.data_point, Relation):
                 relations.append(label.data_point)
         return relations
 
-    def get_spans(self, type: str) -> List[Span]:
+    def get_spans(self, label_type: Optional[str] = None) -> List[Span]:
         spans: List[Span] = []
         for potential_span in self._known_spans.values():
-            if isinstance(potential_span, Span) and potential_span.has_label(type):
+            if isinstance(potential_span, Span) and (label_type is None or potential_span.has_label(label_type)):
                 spans.append(potential_span)
         return sorted(spans)
 
@@ -937,16 +962,13 @@ def to_original_text(self) -> str:
         ).strip()
 
     def to_dict(self, tag_type: Optional[str] = None):
-        labels = []
-
-        if tag_type:
-            labels = [label.to_dict() for label in self.get_labels(tag_type)]
-            return {"text": self.to_original_text(), tag_type: labels}
-
-        if self.labels:
-            labels = [label.to_dict() for label in self.labels]
-
-        return {"text": self.to_original_text(), "all labels": labels}
+        return {
+            "text": self.to_original_text(),
+            "labels": [label.to_dict() for label in self.get_labels(tag_type) if label.data_point is self],
+            "entities": [span.to_dict(tag_type) for span in self.get_spans(tag_type)],
+            "relations": [relation.to_dict(tag_type) for relation in self.get_relations(tag_type)],
+            "tokens": [token.to_dict(tag_type) for token in self.tokens],
+        }
 
     def get_span(self, start: int, stop: int):
         span_slice = slice(start, stop)

flair/models/relation_classifier_model.py

@@ -347,7 +347,7 @@ def _valid_entities(self, sentence: Sentence) -> Iterator[_Entity]:
         :return: Valid entities as `_Entity`
         """
         for label_type, valid_labels in self.entity_label_types.items():
-            for entity_span in sentence.get_spans(type=label_type):
+            for entity_span in sentence.get_spans(label_type=label_type):
                 entity_label: Label = entity_span.get_label(label_type=label_type)
 
                 # Only use entities labelled with the specified labels for each label type