def test_load(self):
wordnet = WordNetModel()
wordnet.load("slowosiec-graph-polish.bin")
n_vertices = len(list(wordnet.g.vertices()))
n_edges = len(list(wordnet.g.edges()))
self.assertGreater(n_vertices, 0)
self.assertGreater(n_edges, 0)
def test_create_graph(self):
wordnet = WordNetModel()
wordnet.create_graph()
n_vertices = len(list(wordnet.g.vertices()))
n_edges = len(list(wordnet.g.edges()))
self.assertGreater(n_vertices, 0)
self.assertGreater(n_edges, 0)
def main():
test_dataset = data_loader.load_simlex_dataset()
print(test_dataset)
results = pd.DataFrame(columns=results_columns)
for model_name in models.keys():
model = WordNetModel()
# wordnet.load("slowosiec-graph-2019-11-07-t17-45.bin")
model.load(f"{model_name}.bin")
values_to_check, missed_words_array, missed_words, unique_missed_words = get_values_to_check(
model, test_dataset)
for similarity_function in similarity_functions.keys():
run_results, run_results_pd = test_model(
model=model,
values_to_check=values_to_check,
similarity_function=similarity_function)
run_results_pd.to_csv(
f'{OUTPUT_PATH}/counted_results/{model_name}-{similarity_function}'
)
for metric in metrics:
model_result = {
"metric":
metric,
"similarity_func":
reverted_similarity_functions[similarity_function],
"similarity_cor":
metrics[metric](run_results[0, :], run_results[2, :]),
"relatedness_cor":
metrics[metric](run_results[1, :], run_results[2, :]),
"model_name":
model_name,
"missed_words_array":
missed_words_array,
"missed_words":
missed_words,
"unique_missed_words":
unique_missed_words,
}
model_results_df = pd.DataFrame(data=[model_result],
columns=results_columns)
results = pd.concat([results, model_results_df])
results = results.reset_index()
results = results.drop(columns=["index"])
print(results)
results.to_csv(f"{OUTPUT_PATH}/wordnet_results.csv")
def test_save_mixed_graph(self):
"""
Used to create binary version of mixed type graph.
:return:
"""
wordnet = WordNetModel()
wordnet.create_graph(is_polish=False)
wordnet.create_lexical_graph(is_polish=False)
file_name = "slowosiec-graph-hiperonim-lexical-selected-connected-by-top-sort"
wordnet.save(file_name)
file_path = os.path.join(DATA_PATH,
os.path.join("models", file_name + ".bin"))
self.assertTrue(os.path.isfile(file_path))
def test_wu_palmer_synonyms(self):
"""
Best are options:
- filtering polish and all types of relations (slowosiec-graph-polish)
- not filtering polish and hiperonim/hiponim relation type (slowosiec-graph-hiponim)
:return:
"""
wordnet = WordNetModel()
wordnet.load("slowosiec-graph-hiperonim-connected-by-top-sort.bin")
synonyms1 = wordnet.synonyms("krzesło", dist_type="WuPalmer")
synonyms2 = wordnet.synonyms("piec", dist_type="WuPalmer")
synonyms3 = wordnet.synonyms("król", dist_type="WuPalmer")
print(synonyms1)
print(synonyms2)
print(synonyms3)
self.assertTrue(synonyms1 is not None)
self.assertTrue(synonyms2 is not None)
self.assertTrue(synonyms3 is not None)
self.assertTrue(type(synonyms1) == list)
self.assertTrue(type(synonyms2) == list)
self.assertTrue(type(synonyms3) == list)
self.assertTrue(type(synonyms1[0]) == str)
self.assertTrue(type(synonyms2[0]) == str)
self.assertTrue(type(synonyms3[0]) == str)
def test__WuPalmer(self):
wordnet = WordNetModel()
wordnet.load("slowosiec-graph-hiperonim-connected-by-top-sort.bin")
dist = wordnet._WuPalmer()
print(dist("łatwy", "męczący"))
def test__count_root_avg_depth(self):
wordnet = WordNetModel()
wordnet.load("slowosiec-graph-hiperonim-connected-by-top-sort.bin")
depth = wordnet._count_root_depth()
print(depth)
self.assertGreater(depth, 0)