def post(self, request, num_task):
"""
Post function used to calculate similarity between models of a word
:param request: request Object
:param num_task: ID of task used
"""
word_form = self.word_class()
word = request.POST['word']
topn = int(request.POST['topn'])
task = Task.objects.get(pk=num_task)
comparable_models = calculate_comparable_models(task)
modelChoice = request.POST['modelChoice'].split("-")
selected_models = [
Word2Vec.load(task.model_set.get(name=modelChoice[0]).model),
Word2Vec.load(task.model_set.get(name=modelChoice[1]).model)
]
models = [Word2Vec.load(model.model) for model in task.model_set.all()]
results = [
(word,
round(lncs2(word, selected_models[0], selected_models[1], topn),
2), round(c_measure(word, selected_models)[0], 2))
]
return render(
request, 'similarity.html', {
'results': results,
'word_form': word_form,
'task': task,
'comparable_models': comparable_models,
'word': word,
'topn': topn,
'modelChoice': request.POST['modelChoice'],
})
def fitness(threshold: float, topn: int, t: float):
similarity = 3
# Task 1 - Binary Classification
predictions = []
for word in binary_truth[:, 0]:
if similarity == 0: # cosine similarity
prediction = (0 if
1 - cosine(model1[word], model2[word]) >= threshold
else 1)
elif similarity == 1: # lncs2 similarity
prediction = (0 if lncs2(word, model1, model2, topn) >= threshold
else 1)
elif similarity == 2: # intersection of nearest neighbours
prediction = (1 if intersection_nn(word, model1, model2, topn) >=
threshold else 0)
elif similarity == 3:
prediction = (0 if moving_lncs2(word, model1, model2, topn, t) >=
threshold else 1)
predictions.append(prediction)
result_accu = accuracy_score(
binary_truth[:, 1].astype(int),
np.array(predictions),
)
result_f1 = f1_score(
binary_truth[:, 1].astype(int),
np.array(predictions),
)
logger.info("F1: " + str(result_f1))
logger.info("PRE: " + str(
precision_score(
binary_truth[:, 1].astype(int),
np.array(predictions),
)))
logger.info("REC: " + str(
recall_score(
binary_truth[:, 1].astype(int),
np.array(predictions),
)))
logger.info("ACC: " + str(result_accu))
return result_accu
# Load binary truths
binary_truth = numpy.loadtxt(
"./data/"
+ lang
+ "/semeval2020_ulscd_"
+ lang[:3]
+ "/truth/binary.txt",
dtype=str,
delimiter="\t",
)
# Task 1 - Binary Classification
predictions = []
for word in binary_truth[:, 0]:
prediction = (
0
if lncs2(word, model1, model2, 10) >= config["THRESHOLD"]
else 1
)
predictions.append(prediction)
logger.info("CLassification score for " + lang)
logger.info(
"\n"
+ classification_report(
binary_truth[:, 1].astype(float),
numpy.array(predictions),
target_names=["class 0 (stable)", "class 1 (change)"],
)
)
# Load scores truths
score_truth = numpy.loadtxt(
"./data/"
CURRENT_EXP_DIR.split("_")[0] + "_0" + "/model/" + lang +
"/corpus1.model")
model2 = Word2Vec.load(
CURRENT_EXP_DIR.split("_")[0] + "_0" + "/model/" + lang +
"/corpus2.model")
# Load binary truths
binary_truth = numpy.loadtxt(
"./data/" + lang + "/semeval2020_ulscd_" + lang[:3] +
"/targets.txt",
dtype=str,
delimiter="\t",
)
# Task 1 - Binary Classification
predictions = []
for word in binary_truth:
prediction = (0 if lncs2(word, model1, model2, 10) >=
config["THRESHOLD"] else 1)
predictions.append([word, str(prediction)])
numpy.savetxt(
CURRENT_EXP_DIR + "/res/answer/task1/" + lang + ".txt",
numpy.array(predictions),
fmt="%s",
delimiter="\t",
)
# Load scores truths
score_truth = numpy.loadtxt(
"./data/" + lang + "/semeval2020_ulscd_" + lang[:3] +
"/targets.txt",
dtype=str,
delimiter="\t",
)