def test_tfidf_asm_similarity(self):
from ir_log import IRLog
from ir_config import IRConfig
from ir_tfidf import IRTFIDF
IRLog.get_instance().start_log()
IRConfig.get_instance().load('../data/test/bug_test.cfg')
vec_a = {'firefox':1, 'chrome':1}
vec_b = {'firefox':1, 'chrome':1, 'ie':1}
vec_c = {'firefox':1, 'windows':1, 'linux':1}
delta = 0.0001
assert abs(1.0 - IRTFIDF.tfidf_asm_similarity(vec_a, vec_b)) < delta
assert abs(0.5 - IRTFIDF.tfidf_asm_similarity(vec_a, vec_c)) < delta
assert IRTFIDF.tfidf_asm_similarity(vec_a, vec_b) > \
IRTFIDF.tfidf_asm_similarity(vec_a, vec_b, None, ['ie'], 100)
def similarity_with(self, other_report):
"""
Returns:
[float, float, float, float], [total score, summary, \
description, stacktrace]
"""
from ir_config import IRConfig
from ir_tfidf import IRTFIDF
from ir_gnome_st_tools import IRSTTools
summary_ratio = IRConfig.get_instance().get_float('bug_summary_ratio')
description_ratio = IRConfig.get_instance().get_float(
'bug_description_ratio')
stacktrace_ratio = IRConfig.get_instance().get_float(
'bug_stacktrace_ratio')
summary_tfidf_a, description_tfidf_a = \
self.get_summary_and_description_tfidf()
summary_tfidf_b, description_tfidf_b = \
other_report.get_summary_and_description_tfidf()
tfidf_algorithm = IRConfig.get_instance().get('tfidf_algorithm')
stacktrace_algorithm = IRConfig.get_instance().get(
'stacktrace_algorithm')
if tfidf_algorithm == 'tfidf':
summary_similarity = IRTFIDF.tfidf_similarity(
summary_tfidf_a, summary_tfidf_b)
description_similarity = IRTFIDF.tfidf_similarity(
description_tfidf_a, description_tfidf_b)
elif tfidf_algorithm == 'bidf':
summary_squared_length, description_squared_length = \
self.get_summary_and_description_tfidf_squared_length()
summary_similarity = IRTFIDF.tfidf_asm_similarity(
summary_tfidf_a, summary_tfidf_b, summary_squared_length)
description_similarity = IRTFIDF.tfidf_asm_similarity(
description_tfidf_a, description_tfidf_b,
description_squared_length, self.__penalty_terms)
if self.__stacktrace is None or \
self.__stacktrace.__len__() == 0 or \
self.__stacktrace[0].__len__() == 0:
stacktrace_similarity = 1.0
else:
stacktrace_similarity = IRSTTools.compare_stackinfo(
self.get_stacktrace(), other_report.get_stacktrace(),
stacktrace_algorithm)
scoring_strategy = IRConfig.get_instance().get('scoring_strategy',
'heuristic')
if scoring_strategy == 'weighted':
score = self.__weighted_scoring(summary_similarity,
description_similarity,
stacktrace_similarity)
elif scoring_strategy == 'heuristic':
score = self.__heuristic_scoring(summary_similarity,
description_similarity,
stacktrace_similarity)
elif scoring_strategy == 'distweighted':
score = self.__distweighted_scoring(summary_similarity,
description_similarity,
stacktrace_similarity)
else:
assert False, 'invalid scoring strategy'
return [
score, summary_similarity, description_similarity,
stacktrace_similarity
]
def similarity_with(self, other_report):
"""
Returns:
[float, float, float, float], [total score, summary, \
description, stacktrace]
"""
from ir_config import IRConfig
from ir_tfidf import IRTFIDF
from ir_gnome_st_tools import IRSTTools
summary_ratio = IRConfig.get_instance().get_float('bug_summary_ratio')
description_ratio = IRConfig.get_instance().get_float('bug_description_ratio')
stacktrace_ratio = IRConfig.get_instance().get_float('bug_stacktrace_ratio')
summary_tfidf_a, description_tfidf_a = \
self.get_summary_and_description_tfidf()
summary_tfidf_b, description_tfidf_b = \
other_report.get_summary_and_description_tfidf()
tfidf_algorithm = IRConfig.get_instance().get('tfidf_algorithm')
stacktrace_algorithm = IRConfig.get_instance().get('stacktrace_algorithm')
if tfidf_algorithm == 'tfidf':
summary_similarity = IRTFIDF.tfidf_similarity(
summary_tfidf_a, summary_tfidf_b)
description_similarity = IRTFIDF.tfidf_similarity(
description_tfidf_a, description_tfidf_b)
elif tfidf_algorithm == 'bidf':
summary_squared_length, description_squared_length = \
self.get_summary_and_description_tfidf_squared_length()
summary_similarity = IRTFIDF.tfidf_asm_similarity(
summary_tfidf_a, summary_tfidf_b, summary_squared_length)
description_similarity = IRTFIDF.tfidf_asm_similarity(
description_tfidf_a, description_tfidf_b,
description_squared_length,
self.__penalty_terms)
if self.__stacktrace is None or \
self.__stacktrace.__len__() == 0 or \
self.__stacktrace[0].__len__() == 0:
stacktrace_similarity = 1.0
else:
stacktrace_similarity = IRSTTools.compare_stackinfo(
self.get_stacktrace(), other_report.get_stacktrace(),
stacktrace_algorithm)
scoring_strategy = IRConfig.get_instance().get('scoring_strategy',
'heuristic')
if scoring_strategy == 'weighted':
score = self.__weighted_scoring(summary_similarity,
description_similarity, stacktrace_similarity)
elif scoring_strategy == 'heuristic':
score = self.__heuristic_scoring(summary_similarity,
description_similarity, stacktrace_similarity)
elif scoring_strategy == 'distweighted':
score = self.__distweighted_scoring(summary_similarity,
description_similarity, stacktrace_similarity)
else:
assert False, 'invalid scoring strategy'
return [score,
summary_similarity,
description_similarity,
stacktrace_similarity]