def get_similarity(text_1, text_2):
#text_1 and text_2 are xml data that uses spans to seperate boundaries
#e.g. BOSTON, MA ... <span class="highlighted" id="634541">Steven L.
#Davis pled guilty yesterday to federal charges that he stole and disclosed trade secrets of The Gillette Company</span>.
if text_1 == '' or text_2 == '':
return 'Error Text Input Is Empty'
else:
xml_soup_1 = BeautifulSoup(text_1)
xml_soup_2 = BeautifulSoup(text_2)
xml_soup_1 = remove_html_tags(xml_soup_1)
xml_soup_2 = remove_html_tags(xml_soup_2)
segements_1 = get_segements(xml_soup_1)
segements_2 = get_segements(xml_soup_2)
seg_check = check_segment_length(segements_1, segements_2)
if not seg_check:
return 'Error Source Text Was Different'
masses_1 = segeval.convert_positions_to_masses(segements_1)
masses_2 = segeval.convert_positions_to_masses(segements_2)
ss = segeval.segmentation_similarity(masses_1, masses_2)
ss = float(ss)
pk = segeval.pk(masses_1, masses_2)
pk = 1 - float(pk)
win_diff = segeval.window_diff(masses_1, masses_2)
win_diff = 1 - float(win_diff)
return ss, pk, win_diff
def evaluateSegments(reference, hypothesis):
ref, hyp = __initialization(reference, hypothesis)
score=np.array([__getscores(reference,hypothesis)[2],\
float(segeval.pk(ref, hyp)),\
float(segeval.window_diff(ref, hyp)),\
float(segeval.boundary_similarity(ref, hyp)),\
float(segeval.segmentation_similarity(ref, hyp))])
# Return pk, windiff, boundary_sim, segmentation_sim and F_1 score.
return score
def evaluate(self, batch, preds, sent=True, word=True):
""" For a given batch and its corresponding preds, get metrics
batch: Batch instance
preds: list
Usage:
>> from loader import *
>> from modules import *
>>
>> model = TextSeg(lstm_dim=200, score_dim=200, bidir=True, num_layers=2)
>> trainer = Trainer(model=model,
train_dir='../data/wiki_727/train',
val_dir='../data/wiki_50/test',
batch_size=10,
lr=1e-3)
>> evalu = Metrics()
>>
>> batch = sample_and_batch(trainer.train_dir, trainer.batch_size, TRAIN=True)
>> preds = trainer.predict_batch(batch)
>> evalu(batch, preds)
"""
metric_dict = {}
assert (sent or
word), 'Missing: choose sent- and / or word-level evaluation.'
# Word level
if word:
w_true, w_pred = self._word(batch, preds)
metric_dict['w_pk'] = seg.pk(w_pred, w_true)
metric_dict['w_wd'] = seg.window_diff(w_pred,
w_true,
lamprier_et_al_2007_fix=True)
metric_dict['w_ss'] = seg.segmentation_similarity(w_pred, w_true)
metric_dict['w_bs'] = seg.boundary_similarity(w_pred, w_true)
w_confusion = seg.boundary_confusion_matrix(w_pred, w_true)
metric_dict['w_precision'] = seg.precision(w_confusion)
metric_dict['w_recall'] = seg.recall(w_confusion)
metric_dict['w_f1'] = seg.fmeasure(w_confusion)
# Sentence level
if sent:
s_true, s_pred = self._sent(batch, preds)
metric_dict['s_pk'] = seg.pk(s_pred, s_true)
metric_dict['s_wd'] = seg.window_diff(s_pred,
s_true,
lamprier_et_al_2007_fix=True)
metric_dict['s_ss'] = seg.segmentation_similarity(s_pred, s_true)
metric_dict['s_bs'] = seg.boundary_similarity(s_pred, s_true)
s_confusion = seg.boundary_confusion_matrix(s_pred, s_true)
metric_dict['s_precision'] = seg.precision(s_confusion)
metric_dict['s_recall'] = seg.recall(s_confusion)
metric_dict['s_f1'] = seg.fmeasure(s_confusion)
return metric_dict
def get_Segmentation_similarity(reference, hypothesis):
ref, hyp = __initialization(reference, hypothesis)
# Evaluate algorithm using S (segmentation similarity) metric
return segeval.segmentation_similarity(ref, hyp)