def test_valid_input(self): self.assertEqual(cosine(['data', 'science'], ['data']), 1.0 / (math.sqrt(2) * math.sqrt(1))) self.assertEqual(cosine(['data', 'science'], ['science', 'good']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual(cosine([], ['data']), 0.0) self.assertEqual(cosine(['data', 'data', 'science'], ['data', 'management']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual(cosine(['data', 'management'], ['data', 'data', 'science']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual(cosine([], []), 1.0) self.assertEqual(cosine(set([]), set([])), 1.0) self.assertEqual(cosine({1, 1, 2, 3, 4}, {2, 3, 4, 5, 6, 7, 7, 8}), 3.0 / (math.sqrt(4) * math.sqrt(7)))
def test_valid_input(self): self.assertEqual(cosine(['data', 'science'], ['data']), 1.0 / (math.sqrt(2) * math.sqrt(1))) self.assertEqual(cosine(['data', 'science'], ['science', 'good']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual(cosine([], ['data']), 0.0) self.assertEqual( cosine(['data', 'data', 'science'], ['data', 'management']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual( cosine(['data', 'management'], ['data', 'data', 'science']), 1.0 / (math.sqrt(2) * math.sqrt(2))) self.assertEqual(cosine([], []), 1.0) self.assertEqual(cosine(set([]), set([])), 1.0) self.assertEqual(cosine({1, 1, 2, 3, 4}, {2, 3, 4, 5, 6, 7, 7, 8}), 3.0 / (math.sqrt(4) * math.sqrt(7)))
def test_invalid_input3(self): cosine(None, None)
def test_invalid_input2(self): cosine(None, ['b'])
def test_invalid_input4(self): cosine(['a'], None)
def test_invalid_input1(self): cosine(1, 1)
def generate_feature(file_name): lines = stage3_helper.read_file(file_name) features = [] labels = [] all_names = [] for line in lines: json1, json2, label = stage3_helper.read_jsons_label_from_line(line) string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_name) all_names.append(tokenizers.whitespace(string1)) all_names.append(tokenizers.whitespace(string2)) for line in lines: json1, json2, label = stage3_helper.read_jsons_label_from_line(line) feature = [] # TODO: Add more features and optimize features. # product_type string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_type) string1 = string1.lower() string2 = string2.lower() feature.append( simfunctions.jaccard(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.jaro_winkler(string1, string2, prefix_weight=0.1)) feature.append( simfunctions.jaro( tokenizers.whitespace(string1)[0], tokenizers.whitespace(string2)[0])) # if len(string1) == len(string2): # feature.append(simfunctions.hamming_distance(string1, string2)) # else: # feature.append(5) feature.append( simfunctions.cosine(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.overlap_coefficient(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.monge_elkan(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.tfidf(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append(len(string1)) feature.append(len(string2)) feature.append(len(string1) - len(string2)) feature.append(len(tokenizers.whitespace(string1))) feature.append(len(tokenizers.whitespace(string2))) # product_name string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_name) string1 = string1.lower() string2 = string2.lower() feature.append( simfunctions.jaccard(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.jaro_winkler(string1, string2, prefix_weight=0.1)) if len(string1) == len(string2): feature.append(simfunctions.hamming_distance(string1, string2)) else: feature.append(5) feature.append( simfunctions.cosine(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.overlap_coefficient(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.monge_elkan(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.tfidf(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append(len(string1)) feature.append(len(string2)) feature.append(len(string1) - len(string2)) feature.append( simfunctions.jaro( tokenizers.whitespace(string1)[0], tokenizers.whitespace(string2)[0])) feature.append(len(tokenizers.whitespace(string1))) feature.append(len(tokenizers.whitespace(string2))) # product_segment string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_segment) string1 = string1.lower() string2 = string2.lower() feature.append( simfunctions.jaccard(tokenizers.qgram(string1, 3), tokenizers.qgram(string2, 3))) feature.append( simfunctions.jaro_winkler(string1, string2, prefix_weight=0.1)) # if len(string1) == len(string2): # feature.append(simfunctions.hamming_distance(string1, string2)) # else: # feature.append(5) feature.append( simfunctions.cosine(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.overlap_coefficient(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.monge_elkan(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.tfidf(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.jaro( tokenizers.whitespace(string1)[0], tokenizers.whitespace(string2)[0])) feature.append(len(string1)) feature.append(len(string2)) feature.append(len(string1) - len(string2)) feature.append(len(tokenizers.whitespace(string1))) feature.append(len(tokenizers.whitespace(string2))) # product_long_description string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_long_description) if string1 is None or string2 is None: feature.append(0.5) feature.append(0) feature.append(0) feature.append(0) feature.append(0) # feature.append(0) # feature.append(0) # feature.append(0) # feature.append(0) else: string1 = string1.lower() string2 = string2.lower() string1 = stage3_helper.cleanhtml(string1) string2 = stage3_helper.cleanhtml(string2) string1 = stage3_helper.clean_stop_word(string1) string2 = stage3_helper.clean_stop_word(string2) feature.append( simfunctions.jaccard(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) # feature.append(simfunctions.jaro_winkler(string1, string2, prefix_weight=0.1)) feature.append( simfunctions.overlap_coefficient( tokenizers.whitespace(string1), tokenizers.whitespace(string2))) # feature.append(simfunctions.monge_elkan(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) # feature.append(simfunctions.tfidf(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append(len(string1)) feature.append(len(string2)) feature.append(len(string1) - len(string2)) # product_brand string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_brand) string1_name, string2_name = stage3_helper.get_attribute_from_jsons( json1, json2, product_name) if string1 is None or string1 == '': string1 = get_predict_brand(string1_name) if string2 is None or string2 == '': string2 = get_predict_brand(string2_name) if string1 is None or string2 is None: feature.append(0) feature.append(0) feature.append(0) feature.append(0) feature.append(0) feature.append(0) feature.append(0) feature.append(0) else: feature.append( simfunctions.jaccard(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.jaro_winkler(string1, string2, prefix_weight=0.1)) feature.append( simfunctions.overlap_coefficient( tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.monge_elkan(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append( simfunctions.tfidf(tokenizers.whitespace(string1), tokenizers.whitespace(string2))) feature.append(len(string1)) feature.append(len(string2)) feature.append(len(string1) - len(string2)) #feature.append(simfunctions.jaro(tokenizers.whitespace(string1)[0], tokenizers.whitespace(string2)[0])) # Contains similar model names. string1, string2 = stage3_helper.get_attribute_from_jsons( json1, json2, product_name) string1 = string1.lower() string2 = string2.lower() model_strs1 = stage3_helper.find_model_str(string1) model_strs2 = stage3_helper.find_model_str(string2) # share_model_str = False # for model in model_strs1: # if model.lower() in string2.lower(): # share_model_str = True # for model in model_strs2: # if model.lower() in string1.lower(): # share_model_str = True # if share_model_str: # feature.append(1) # else: # feature.append(0) if len(model_strs1) > 0 and len(model_strs2) > 0: feature.append(simfunctions.jaccard(model_strs1, model_strs2)) else: feature.append(0.5) feature.append(len(model_strs1)) feature.append(len(model_strs2)) feature.append(len(model_strs1) - len(string2)) # Other features. common = 0 common_score = 0.0 for item in json1: if item in json2: common += 1 common_score += simfunctions.jaccard( tokenizers.whitespace(json1[item][0]), tokenizers.whitespace(json2[item][0])) common_score = common_score / common feature.append(len(json1)) feature.append(len(json2)) feature.append(len(json1) - len(json2)) feature.append(common) feature.append(common_score) feature.append(len(json.dumps(json1))) feature.append(len(json.dumps(json2))) feature.append(len(json.dumps(json1)) - len(json.dumps(json2))) feature.append( simfunctions.jaccard(tokenizers.whitespace(json.dumps(json1)), tokenizers.whitespace(json.dumps(json2)))) # Add one feature and label. features.append(feature) labels.append(stage3_helper.get_01_from_label(label)) return features, labels, lines
from py_stringmatching import simfunctions, tokenizers prodNames = [] cosineMeasure = [] with open('set_X.txt', 'r') as f: for line in f: line = unicode(line, errors='ignore') # For character which are not utf-8 data = line.strip().split('?') pairId = data[0] prod1_id = data[1] if (data[2]): prod1_json = json.loads(data[2]) else: prod1_json = dict() prod2_id = data[3] if (data[4]): prod2_json = json.loads(data[4]) else: prod2_json = dict() label = data[5] prodNames.append( (prod1_json['Product Name'][0], prod2_json['Product Name'][0])) f.close() for pair in prodNames: measure = simfunctions.cosine(tokenizers.whitespace(pair[0]), tokenizers.whitespace(pair[1])) cosineMeasure.append(measure) print cosineMeasure
def time_medium_large_wi_rep(self): simfunctions.cosine(_med_num_tokens_wo_rep, _large_num_tokens_wo_rep)
def time_small_large_wi_rep(self): simfunctions.cosine(_small_num_tokens_wi_rep, _large_num_tokens_wi_rep)
def time_small_medium_wi_rep(self): simfunctions.cosine(_small_num_tokens_wi_rep, _med_num_tokens_wi_rep)
def time_small_small_wo_rep(self): simfunctions.cosine(_small_num_tokens_wo_rep, _small_num_tokens_wo_rep)