#398 related solution in 0.22.1.

arekit/contrib/source/sentinerel/entities.py

@@ -35,10 +35,6 @@ def __keep_entity(self, entity):
 
     @classmethod
     def read_collection(cls, filename, version, entities_to_ignore=None):
-        """ You may adopt the following list of entities to ignore by default.
-                entities_to_ignore = ["EFFECT_NEG", "EFFECT_POS", "ARGUMENT_NEG", "ARGUMENT_POS", "EVENT"]
-            We setup the latter as a part of the side-project and the related experiments.
-        """
         assert(isinstance(filename, str))
 
         # Since this dataset does not provide the synonyms collection by default,

arekit/contrib/source/sentinerel/reader.py

@@ -19,15 +19,22 @@ def read_text_relations(filename, version):
             version=version)
 
     @staticmethod
-    def read_document(filename, doc_id):
+    def read_document(filename, doc_id, entities_to_ignore=None):
         assert(isinstance(filename, str))
         assert(isinstance(doc_id, int))
 
         def file_to_doc(input_file):
             sentences = BratDocumentSentencesReader.from_file(input_file=input_file, entities=entities)
             return BratNews(doc_id=doc_id, sentences=sentences, text_relations=text_relations)
 
-        entities = SentiNerelEntityCollection.read_collection(filename=filename, version=SentiNerelVersions.V1)
+        # TODO. #398 issue -- in some cases entities might be nested. Therefore we limit the set
+        # TODO. of the potential named entities.
+        eti = ["EFFECT_NEG", "EFFECT_POS", "ARGUMENT_NEG", "ARGUMENT_POS", "EVENT"] \
+            if entities_to_ignore is None else entities_to_ignore
+
+        entities = SentiNerelEntityCollection.read_collection(
+            filename=filename, version=SentiNerelVersions.V1,
+            entities_to_ignore=eti)
         text_relations = SentiNerelDocReader.read_text_relations(filename=filename, version=SentiNerelVersions.V1)
 
         return SentiNerelIOUtils.read_from_zip(