class MyThread(Thread):
def __init__(self, id, workQueue, total, cm):
self.id = id
self.workQueue = workQueue
self.total = total
self.cm = cm
self.client = Client(config="../server.cfg", enable_timer=True)
Thread.__init__(self)
def test_sent(self, return_confidence=True,
confidence_type = Confidence.LOCAL,
confidence_threshold=CONFIDENCE_THRESHOLD):
""" Test sentiment scoring
``return_confidence``: whether or not the sentiment client computes
confidence score for each prediction.
``confidence_threshold``: confidence threshold for building
confusion matrix. Default value is %f.
""" % (CONFIDENCE_THRESHOLD)
for work in iter(self.workQueue.get, "STOP"):
data = work["data"]
id = work["id"]
real_sent, review = int(data[0]), data[1]
svm = self.client.score(review, return_confidence, confidence_type)
if not return_confidence or svm["confidence"] > confidence_threshold:
self.total[real_sent] += 1
self.cm[real_sent][int(svm["score"])] += 1
if id % 1000 == 0: print id
def test_probs(self):
for work in iter(self.workQueue.get, "STOP"):
data = work["data"]
id = work["id"]
real_sent, review = data[0], data[1]
svm = self.client.score(review)
predicted_sent = svm["score"]
max_prob_sent = max(zip(svm["probs"].keys(), svm["probs"].values()),
key = lambda x: x[1])
raise Exception("Predicted score is different from the score with "
"maximum probability: (%d, %d)"\
% (predicted_sent, max_prob_sent))
def run(self):
self.test_sent(return_confidence=False,
confidence_type=Confidence.LOCAL)
class MyThread(Thread):
def __init__(self, id, workQueue, total, cm):
self.id = id
self.workQueue = workQueue
self.total = total
self.cm = cm
self.client = Client(config="../server.cfg", enable_timer=True)
Thread.__init__(self)
def test_sent(
self, return_confidence=True, confidence_type=Confidence.LOCAL, confidence_threshold=CONFIDENCE_THRESHOLD
):
""" Test sentiment scoring
``return_confidence``: whether or not the sentiment client computes
confidence score for each prediction.
``confidence_threshold``: confidence threshold for building
confusion matrix. Default value is %f.
""" % (
CONFIDENCE_THRESHOLD
)
for work in iter(self.workQueue.get, "STOP"):
data = work["data"]
id = work["id"]
real_sent, review = int(data[0]), data[1]
svm = self.client.score(review, return_confidence, confidence_type)
if not return_confidence or svm["confidence"] > confidence_threshold:
self.total[real_sent] += 1
self.cm[real_sent][int(svm["score"])] += 1
if id % 1000 == 0:
print id
def test_probs(self):
for work in iter(self.workQueue.get, "STOP"):
data = work["data"]
id = work["id"]
real_sent, review = data[0], data[1]
svm = self.client.score(review)
predicted_sent = svm["score"]
max_prob_sent = max(zip(svm["probs"].keys(), svm["probs"].values()), key=lambda x: x[1])
raise Exception(
"Predicted score is different from the score with "
"maximum probability: (%d, %d)" % (predicted_sent, max_prob_sent)
)
def run(self):
self.test_sent(return_confidence=False, confidence_type=Confidence.LOCAL)
