def _display(self):
if self._mode == 'basic':
rows = self._cursor.fetchall()
if not rows:
return ''
newrows = []
for row in rows:
src, tgt, pe2e1, rel, pivnum, piv, dist = row
pe2e1 = math.exp(-float(pe2e1))
pe2e1str = '{:>6.4}'.format(
pe2e1) if pe2e1 < 0.0001 else '{:0<6.4f}'.format(pe2e1)
rel_or_path = para_wn.get_relation_name(rel)
if rel_or_path == 'undefined relation':
rel_or_path = 'WN distance=' + str(
dist) if dist >= 0 else 'not connected in WN'
newrows.append([
src.encode('utf-8'),
tgt.encode('utf-8'), pe2e1str, rel_or_path, pivnum,
piv.encode('utf-8'), dist
])
srclens = map(len, map(operator.itemgetter(0), newrows))
trglens = map(len, map(operator.itemgetter(1), newrows))
probstrlens = map(len, map(operator.itemgetter(2), newrows))
relationlens = map(len, map(operator.itemgetter(3), newrows))
maxsrclen, maxtrglen, maxproblen, maxrellen = max(srclens), max(
trglens), max(probstrlens), max(relationlens)
return self._format_display(newrows, self._interactive, maxsrclen,
maxtrglen, maxproblen, maxrellen)
# display the count results
else:
rows = self._cursor.fetchall()
if not rows:
return ''
res = ['']
rows = sorted(rows)
if self._group_by == 'relation':
rows = [(para_wn.get_relation_name(group), count)
for (group, count) in rows]
elif self._group_by == 'samepos':
rows = [(self._POS_IDX_TO_VALUES[group], count)
for (group, count) in rows]
else:
rows = [(str(group), count) for (group, count) in rows]
grouplens = map(len, map(operator.itemgetter(0), rows))
maxgrouplen = max(grouplens)
for group, count in rows:
if group == '':
res.append(str(count))
else:
res.append('{}\t{}'.format(group.rjust(maxgrouplen),
count))
res.append('')
return '\n'.join(res)
def _display(self):
if self._mode == 'basic':
rows = self._cursor.fetchall()
if not rows:
return ''
newrows = []
for row in rows:
src, tgt, pe2e1, rel, pivnum, piv, dist = row
pe2e1 = math.exp(-float(pe2e1))
pe2e1str = '{:>6.4}'.format(pe2e1) if pe2e1 < 0.0001 else '{:0<6.4f}'.format(pe2e1)
rel_or_path = para_wn.get_relation_name(rel)
if rel_or_path == 'undefined relation':
rel_or_path = 'WN distance=' + str(dist) if dist >= 0 else 'not connected in WN'
newrows.append([src.encode('utf-8'), tgt.encode('utf-8'), pe2e1str, rel_or_path, pivnum, piv.encode('utf-8'), dist])
srclens = map(len, map(operator.itemgetter(0), newrows))
trglens = map(len, map(operator.itemgetter(1), newrows))
probstrlens = map(len, map(operator.itemgetter(2), newrows))
relationlens = map(len, map(operator.itemgetter(3), newrows))
maxsrclen, maxtrglen, maxproblen, maxrellen = max(srclens), max(trglens), max(probstrlens), max(relationlens)
return self._format_display(newrows, self._interactive, maxsrclen, maxtrglen, maxproblen, maxrellen)
# display the count results
else:
rows = self._cursor.fetchall()
if not rows:
return ''
res = ['']
rows = sorted(rows)
if self._group_by == 'relation':
rows = [(para_wn.get_relation_name(group), count) for (group, count) in rows]
elif self._group_by == 'samepos':
rows = [(self._POS_IDX_TO_VALUES[group], count) for (group, count) in rows]
else:
rows = [(str(group), count) for (group, count) in rows]
grouplens = map(len, map(operator.itemgetter(0), rows))
maxgrouplen = max(grouplens)
for group, count in rows:
if group == '':
res.append(str(count))
else:
res.append('{}\t{}'.format(group.rjust(maxgrouplen), count))
res.append('')
return '\n'.join(res)
def analyze_rules(all_rules, percentile_scores):
db_size = len(all_rules)
data = {}
data[whole] = {}
data[whole]['sample'] = []
data[pivots] = {}
data[pivots][whole] = []
data[distances] = {}
data[distances][whole] = {}
data[distances][whole]['values'] = []
data[per_source] = {}
data[per_source][whole] = {}
data[per_source][whole]['tgtnum'] = []
for part in parts:
data[part] = {}
data[pivots][part] = []
data[distances][part] = {}
data[distances][part]['values'] = []
data[per_source][part] = {}
data[per_source][part]['tgtnum'] = []
random_sample_size = max(int(len(all_rules) * 0.03), 1)
random_sample_size = min(random_sample_size, 25)
data[whole]['sample'] = get_rules_sample(all_rules, [15, 40, 60, 85],
random_sample_size)
cnt = 0
ten_percent_len = db_size / 10
w1 = ""
stats_per_source = {}
stats_per_source[whole] = {}
stats_per_source[whole]['tgtnum'] = 0
for part in parts:
stats_per_source[part] = {}
stats_per_source[part]['tgtnum'] = 0
while len(all_rules) > 0:
cnt = cnt + 1
for percent in range(10):
progress = percent * ten_percent_len
if progress == cnt:
sys.stdout.write(str(percent * 10) + "%... ")
sys.stdout.flush()
break
rule = all_rules.pop(0)
# the results must be sorted by source (rule[0]), if reached new source, save and reset statistics per source
if w1 != rule[0]:
# save the current stats_per_source to data
if w1 != "":
#parts + whole
for x in stats_per_source.keys():
#relations + tgtnum
for y in stats_per_source[x].keys():
if y not in data[per_source][x].keys():
data[per_source][x][y] = []
if stats_per_source[x][y] > 0:
data[per_source][x][y].append(
stats_per_source[x][y])
# reset statistics per source
w1 = rule[0]
stats_per_source = {}
stats_per_source[whole] = {}
stats_per_source[whole]['tgtnum'] = 0
for part in parts:
stats_per_source[part] = {}
stats_per_source[part]['tgtnum'] = 0
w2 = rule[1]
prob = get_prob(rule)
part = get_part(prob, percentile_scores)
rel = wn.get_relation_name(rule[3])
pivotnum = int(rule[4])
dist = get_distance(int(rule[6]))
if rel not in data[whole].keys():
data[whole][rel] = 0
# count how many times each of the relations was observed in the whole collection
data[whole][rel] = data[whole][rel] + 1
data[pivots][whole].append(pivotnum)
# increase the count of targets per current source by 1
stats_per_source[whole]['tgtnum'] += 1
if rel not in stats_per_source[whole].keys():
stats_per_source[whole][rel] = 0
#increse by 1 the count of current relation per source
stats_per_source[whole][rel] += 1
# for undefined relations, save WN distance
if rel == undefined:
if dist not in data[distances][whole].keys():
data[distances][whole][dist] = []
# save the rule with that distance
data[distances][whole][dist].append(rule)
#save the distance itself
data[distances][whole]['values'].append(int(rule[6]))
if part in data.keys():
if rel not in data[part].keys():
data[part][rel] = []
data[part][rel].append(rule)
data[pivots][part].append(pivotnum)
if rel == undefined:
#for undefined relations, save WN distance
if dist not in data[distances][part].keys():
data[distances][part][dist] = []
data[distances][part][dist].append(rule)
data[distances][part]['values'].append(int(rule[6]))
# increase the count of targets per current source by 1
stats_per_source[part]['tgtnum'] += 1
if rel not in stats_per_source[part].keys():
stats_per_source[part][rel] = 0
#increse by 1 the count of current relation per source
stats_per_source[part][rel] += 1
print
# save the last stats_per_source to data
for x in stats_per_source.keys():
# parts + whole
for y in stats_per_source[x].keys():
# relations + tgtnum
if y not in data[per_source][x].keys():
data[per_source][x][y] = []
if stats_per_source[x][y] > 0:
data[per_source][x][y].append(stats_per_source[x][y])
return data
def analyze_rules(all_rules, percentile_scores):
db_size = len(all_rules)
data = {}
data[whole] = {}
data[whole]["sample"] = []
data[pivots] = {}
data[pivots][whole] = []
data[distances] = {}
data[distances][whole] = {}
data[distances][whole]["values"] = []
data[per_source] = {}
data[per_source][whole] = {}
data[per_source][whole]["tgtnum"] = []
for part in parts:
data[part] = {}
data[pivots][part] = []
data[distances][part] = {}
data[distances][part]["values"] = []
data[per_source][part] = {}
data[per_source][part]["tgtnum"] = []
random_sample_size = max(int(len(all_rules) * 0.03), 1)
random_sample_size = min(random_sample_size, 25)
data[whole]["sample"] = get_rules_sample(all_rules, [15, 40, 60, 85], random_sample_size)
cnt = 0
ten_percent_len = db_size / 10
w1 = ""
stats_per_source = {}
stats_per_source[whole] = {}
stats_per_source[whole]["tgtnum"] = 0
for part in parts:
stats_per_source[part] = {}
stats_per_source[part]["tgtnum"] = 0
while len(all_rules) > 0:
cnt = cnt + 1
for percent in range(10):
progress = percent * ten_percent_len
if progress == cnt:
sys.stdout.write(str(percent * 10) + "%... ")
sys.stdout.flush()
break
rule = all_rules.pop(0)
# the results must be sorted by source (rule[0]), if reached new source, save and reset statistics per source
if w1 != rule[0]:
# save the current stats_per_source to data
if w1 != "":
# parts + whole
for x in stats_per_source.keys():
# relations + tgtnum
for y in stats_per_source[x].keys():
if y not in data[per_source][x].keys():
data[per_source][x][y] = []
if stats_per_source[x][y] > 0:
data[per_source][x][y].append(stats_per_source[x][y])
# reset statistics per source
w1 = rule[0]
stats_per_source = {}
stats_per_source[whole] = {}
stats_per_source[whole]["tgtnum"] = 0
for part in parts:
stats_per_source[part] = {}
stats_per_source[part]["tgtnum"] = 0
w2 = rule[1]
prob = get_prob(rule)
part = get_part(prob, percentile_scores)
rel = wn.get_relation_name(rule[3])
pivotnum = int(rule[4])
dist = get_distance(int(rule[6]))
if rel not in data[whole].keys():
data[whole][rel] = 0
# count how many times each of the relations was observed in the whole collection
data[whole][rel] = data[whole][rel] + 1
data[pivots][whole].append(pivotnum)
# increase the count of targets per current source by 1
stats_per_source[whole]["tgtnum"] += 1
if rel not in stats_per_source[whole].keys():
stats_per_source[whole][rel] = 0
# increse by 1 the count of current relation per source
stats_per_source[whole][rel] += 1
# for undefined relations, save WN distance
if rel == undefined:
if dist not in data[distances][whole].keys():
data[distances][whole][dist] = []
# save the rule with that distance
data[distances][whole][dist].append(rule)
# save the distance itself
data[distances][whole]["values"].append(int(rule[6]))
if part in data.keys():
if rel not in data[part].keys():
data[part][rel] = []
data[part][rel].append(rule)
data[pivots][part].append(pivotnum)
if rel == undefined:
# for undefined relations, save WN distance
if dist not in data[distances][part].keys():
data[distances][part][dist] = []
data[distances][part][dist].append(rule)
data[distances][part]["values"].append(int(rule[6]))
# increase the count of targets per current source by 1
stats_per_source[part]["tgtnum"] += 1
if rel not in stats_per_source[part].keys():
stats_per_source[part][rel] = 0
# increse by 1 the count of current relation per source
stats_per_source[part][rel] += 1
print
# save the last stats_per_source to data
for x in stats_per_source.keys():
# parts + whole
for y in stats_per_source[x].keys():
# relations + tgtnum
if y not in data[per_source][x].keys():
data[per_source][x][y] = []
if stats_per_source[x][y] > 0:
data[per_source][x][y].append(stats_per_source[x][y])
return data