improve tests and switch to upstream flair until flairNLP/flair#3608 …

…is part of a new release

natural_language_processing/tests/conftest.py

@@ -0,0 +1,30 @@
+import os
+import pytest
+import json
+
+from natural_language_processing.roberta_ner import RobertaNER
+from natural_language_processing.flair_ner import FlairNER
+
+
+@pytest.fixture(scope="session")
+def news_items():
+    dir_path = os.path.dirname(os.path.realpath(__file__))
+    story_json = os.path.join(dir_path, "story_list.json")
+    with open(story_json) as f:
+        data = json.load(f)
+    yield [item["content"] for cluster in data for item in cluster["news_items"] if "content" in item]
+
+
+@pytest.fixture(scope="session")
+def flair_analyzer():
+    yield FlairNER()
+
+
+@pytest.fixture(scope="session")
+def roberta_analyzer():
+    yield RobertaNER()
+
+
+@pytest.fixture(scope="session")
+def example_text():
+    yield "This is an example for NER, about the ACME Corporation which is producing Dynamite in Acme City, which is in Australia and run by Mr. Wile E. Coyote"

natural_language_processing/tests/test_analysis.py

@@ -0,0 +1,14 @@
+from natural_language_processing.roberta_ner import RobertaNER
+from natural_language_processing.flair_ner import FlairNER
+
+
+def test_analyze_ner_flair(example_text: str, flair_analyzer: FlairNER):
+    result = flair_analyzer.predict(example_text)
+    expected = {"Australia": "LOC", "Wile E. Coyote": "PER"}
+    assert expected.items() <= result.items()
+
+
+def test_analyze_ner_roberta(example_text: str, roberta_analyzer: RobertaNER):
+    result = roberta_analyzer.predict(example_text)
+    expected = {"ACME Corporation": "ORG", "Acme City": "LOC", "Australia": "LOC", "Dynamite": "MISC"}
+    assert expected.items() <= result.items()

pyproject.toml

@@ -18,7 +18,7 @@ dependencies = [
     "requests",
     "granian",
     "sentencepiece",
-    "flair",  
+    "flair @ git+https://github.com/flairNLP/flair@master",
     "torch",
     "huggingface_hub",
     "pydantic-settings",