# -*- coding: utf-8 -*-
"""
@author: Chenglong Chen <c.chenglong@gmail.com>
@brief: statistical cooccurrence (weighted) features
- TF & normalized TF
- TFIDF & normalized TFIDF
- Okapi BM25
"""
import sys
import string
from collections import defaultdict
import numpy as np
import pandas as pd
import config
from utils import dist_utils, ngram_utils, nlp_utils, np_utils, pkl_utils
from utils import logging_utils, time_utils
from feature_base import BaseEstimator, PairwiseFeatureWrapper
# tune the token pattern to get a better correlation with y_train
# token_pattern = r"(?u)\b\w\w+\b"
# token_pattern = r"\w{1,}"
# token_pattern = r"\w+"
# token_pattern = r"[\w']+"
token_pattern = " " # just split the text into tokens
# ----------------------------- TF ------------------------------------
# StatCooc stands for StatisticalCooccurrence
class StatCoocTF_Ngram(BaseEstimator):
"""Single aggregation features"""
def __init__(self, obs_corpus, target_corpus, ngram, aggregation_mode="",
str_match_threshold=config.STR_MATCH_THRESHOLD):
super().__init__(obs_corpus, target_corpus, aggregation_mode)
self.ngram = ngram
self.ngram_str = ngram_utils._ngram_str_map[self.ngram]
self.str_match_threshold = str_match_threshold
def __name__(self):
if isinstance(self.aggregation_mode, str):
feat_name = "StatCoocTF_%s_%s"%(
self.ngram_str, string.capwords(self.aggregation_mode))
elif isinstance(self.aggregation_mode, list):
feat_name = ["StatCoocTF_%s_%s"%(
self.ngram_str, string.capwords(m)) for m in self.aggregation_mode]
return feat_name
def transform_one(self, obs, target, id):
obs_tokens = nlp_utils._tokenize(obs, token_pattern)
target_tokens = nlp_utils._tokenize(target, token_pattern)
obs_ngrams = ngram_utils._ngrams(obs_tokens, self.ngram)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
val_list = []
for w1 in obs_ngrams:
s = 0.
for w2 in target_ngrams:
if dist_utils._is_str_match(w1, w2, self.str_match_threshold):
s += 1.
val_list.append(s)
if len(val_list) == 0:
val_list = [config.MISSING_VALUE_NUMERIC]
return val_list
# ----------------------------- Normalized TF ------------------------------------
# StatCooc stands for StatisticalCooccurrence
class StatCoocNormTF_Ngram(BaseEstimator):
"""Single aggregation features"""
def __init__(self, obs_corpus, target_corpus, ngram, aggregation_mode="",
str_match_threshold=config.STR_MATCH_THRESHOLD):
super().__init__(obs_corpus, target_corpus, aggregation_mode)
self.ngram = ngram
self.ngram_str = ngram_utils._ngram_str_map[self.ngram]
self.str_match_threshold = str_match_threshold
def __name__(self):
if isinstance(self.aggregation_mode, str):
feat_name = "StatCoocNormTF_%s_%s"%(
self.ngram_str, string.capwords(self.aggregation_mode))
elif isinstance(self.aggregation_mode, list):
feat_name = ["StatCoocNormTF_%s_%s"%(
self.ngram_str, string.capwords(m)) for m in self.aggregation_mode]
return feat_name
def transform_one(self, obs, target, id):
obs_tokens = nlp_utils._tokenize(obs, token_pattern)
target_tokens = nlp_utils._tokenize(target, token_pattern)
obs_ngrams = ngram_utils._ngrams(obs_tokens, self.ngram)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
val_list = []
for w1 in obs_ngrams:
s = 0.
for w2 in target_ngrams:
if dist_utils._is_str_match(w1, w2, self.str_match_threshold):
s += 1.
val_list.append(np_utils._try_divide(s, len(target_ngrams)))
if len(val_list) == 0:
val_list = [config.MISSING_VALUE_NUMERIC]
return val_list
# ------------------------------ TFIDF -----------------------------------
# StatCooc stands for StatisticalCooccurrence
class StatCoocTFIDF_Ngram(BaseEstimator):
"""Single aggregation features"""
def __init__(self, obs_corpus, target_corpus, ngram, aggregation_mode="",
str_match_threshold=config.STR_MATCH_THRESHOLD):
super().__init__(obs_corpus, target_corpus, aggregation_mode)
self.ngram = ngram
self.ngram_str = ngram_utils._ngram_str_map[self.ngram]
self.str_match_threshold = str_match_threshold
self.df_dict = self._get_df_dict()
def __name__(self):
if isinstance(self.aggregation_mode, str):
feat_name = "StatCoocTFIDF_%s_%s"%(
self.ngram_str, string.capwords(self.aggregation_mode))
elif isinstance(self.aggregation_mode, list):
feat_name = ["StatCoocTFIDF_%s_%s"%(
self.ngram_str, string.capwords(m)) for m in self.aggregation_mode]
return feat_name
def _get_df_dict(self):
# smoothing
d = defaultdict(lambda : 1)
for target in self.target_corpus:
target_tokens = nlp_utils._tokenize(target, token_pattern)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
for w in set(target_ngrams):
d[w] += 1
return d
def _get_idf(self, word):
return np.log((self.N - self.df_dict[word] + 0.5)/(self.df_dict[word] + 0.5))
def transform_one(self, obs, target, id):
obs_tokens = nlp_utils._tokenize(obs, token_pattern)
target_tokens = nlp_utils._tokenize(target, token_pattern)
obs_ngrams = ngram_utils._ngrams(obs_tokens, self.ngram)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
val_list = []
for w1 in obs_ngrams:
s = 0.
for w2 in target_ngrams:
if dist_utils._is_str_match(w1, w2, self.str_match_threshold):
s += 1.
val_list.append(s * self._get_idf(w1))
if len(val_list) == 0:
val_list = [config.MISSING_VALUE_NUMERIC]
return val_list
# ------------------------------ Normalized TFIDF -----------------------------------
# StatCooc stands for StatisticalCooccurrence
class StatCoocNormTFIDF_Ngram(BaseEstimator):
"""Single aggregation features"""
def __init__(self, obs_corpus, target_corpus, ngram, aggregation_mode="",
str_match_threshold=config.STR_MATCH_THRESHOLD):
super().__init__(obs_corpus, target_corpus, aggregation_mode)
self.ngram = ngram
self.ngram_str = ngram_utils._ngram_str_map[self.ngram]
self.str_match_threshold = str_match_threshold
self.df_dict = self._get_df_dict()
def __name__(self):
if isinstance(self.aggregation_mode, str):
feat_name = "StatCoocNormTFIDF_%s_%s"%(
self.ngram_str, string.capwords(self.aggregation_mode))
elif isinstance(self.aggregation_mode, list):
feat_name = ["StatCoocNormTFIDF_%s_%s"%(
self.ngram_str, string.capwords(m)) for m in self.aggregation_mode]
return feat_name
def _get_df_dict(self):
# smoothing
d = defaultdict(lambda : 1)
for target in self.target_corpus:
target_tokens = nlp_utils._tokenize(target, token_pattern)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
for w in set(target_ngrams):
d[w] += 1
return d
def _get_idf(self, word):
return np.log((self.N - self.df_dict[word] + 0.5)/(self.df_dict[word] + 0.5))
def transform_one(self, obs, target, id):
obs_tokens = nlp_utils._tokenize(obs, token_pattern)
target_tokens = nlp_utils._tokenize(target, token_pattern)
obs_ngrams = ngram_utils._ngrams(obs_tokens, self.ngram)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
val_list = []
for w1 in obs_ngrams:
s = 0.
for w2 in target_ngrams:
if dist_utils._is_str_match(w1, w2, self.str_match_threshold):
s += 1.
val_list.append(np_utils._try_divide(s, len(target_ngrams)) * self._get_idf(w1))
if len(val_list) == 0:
val_list = [config.MISSING_VALUE_NUMERIC]
return val_list
# ------------------------ BM25 ---------------------------------------------
# StatCooc stands for StatisticalCooccurrence
class StatCoocBM25_Ngram(BaseEstimator):
"""Single aggregation features"""
def __init__(self, obs_corpus, target_corpus, ngram, aggregation_mode="",
str_match_threshold=config.STR_MATCH_THRESHOLD, k1=config.BM25_K1, b=config.BM25_B):
super().__init__(obs_corpus, target_corpus, aggregation_mode)
self.k1 = k1
self.b = b
self.ngram = ngram
self.ngram_str = ngram_utils._ngram_str_map[self.ngram]
self.str_match_threshold = str_match_threshold
self.df_dict = self._get_df_dict()
self.avg_ngram_doc_len = self._get_avg_ngram_doc_len()
def __name__(self):
if isinstance(self.aggregation_mode, str):
feat_name = "StatCoocBM25_%s_%s"%(
self.ngram_str, string.capwords(self.aggregation_mode))
elif isinstance(self.aggregation_mode, list):
feat_name = ["StatCoocBM25_%s_%s"%(
self.ngram_str, string.capwords(m)) for m in self.aggregation_mode]
return feat_name
def _get_df_dict(self):
# smoothing
d = defaultdict(lambda : 1)
for target in self.target_corpus:
target_tokens = nlp_utils._tokenize(target, token_pattern)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
for w in set(target_ngrams):
d[w] += 1
return d
def _get_idf(self, word):
return np.log((self.N - self.df_dict[word] + 0.5)/(self.df_dict[word] + 0.5))
def _get_avg_ngram_doc_len(self):
lst = []
for target in self.target_corpus:
target_tokens = nlp_utils._tokenize(target, token_pattern)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
lst.append(len(target_ngrams))
return np.mean(lst)
def transform_one(self, obs, target, id):
obs_tokens = nlp_utils._tokenize(obs, token_pattern)
target_tokens = nlp_utils._tokenize(target, token_pattern)
obs_ngrams = ngram_utils._ngrams(obs_tokens, self.ngram)
target_ngrams = ngram_utils._ngrams(target_tokens, self.ngram)
K = self.k1 * (1 - self.b + self.b * np_utils._try_divide(len(target_ngrams), self.avg_ngram_doc_len))
val_list = []
for w1 in obs_ngrams:
s = 0.
for w2 in target_ngrams:
if dist_utils._is_str_match(w1, w2, self.str_match_threshold):
s += 1.
bm25 = s * self._get_idf(w1) * np_utils._try_divide(1 + self.k1, s + K)
val_list.append(bm25)
if len(val_list) == 0:
val_list = [config.MISSING_VALUE_NUMERIC]
return val_list
# ---------------------------- Main --------------------------------------
def main(which):
logname = "generate_feature_stat_cooc_tfidf_%s_%s.log"%(which, time_utils._timestamp())
logger = logging_utils._get_logger(config.LOG_DIR, logname)
dfAll = pkl_utils._load(config.ALL_DATA_LEMMATIZED_STEMMED)
generators = []
if which == "tf":
generators.append( StatCoocTF_Ngram )
elif which == "norm_tf":
generators.append( StatCoocNormTF_Ngram )
elif which == "tfidf":
generators.append( StatCoocTFIDF_Ngram )
elif which == "norm_tfidf":
generators.append( StatCoocNormTFIDF_Ngram )
elif which == "bm25":
generators.append( StatCoocBM25_Ngram )
obs_fields_list = []
target_fields_list = []
## query in document
obs_fields_list.append( ["search_term", "search_term_alt", "search_term_auto_corrected"][:1] )
target_fields_list.append( ["product_title", "product_title_product_name", "product_description", "product_attribute", "product_brand", "product_color"] )
## document in query
obs_fields_list.append( ["product_title", "product_title_product_name", "product_description", "product_attribute", "product_brand", "product_color"] )
target_fields_list.append( ["search_term", "search_term_alt", "search_term_auto_corrected"][:1] )
ngrams = [1,2,3,12,123][:3]
aggregation_mode = ["mean", "std", "max", "min", "median"]
for obs_fields, target_fields in zip(obs_fields_list, target_fields_list):
for generator in generators:
for ngram in ngrams:
param_list = [ngram, aggregation_mode]
pf = PairwiseFeatureWrapper(generator, dfAll, obs_fields, target_fields, param_list, config.FEAT_DIR, logger)
pf.go()
obs_fields_list = []
target_fields_list = []
## query in document
obs_fields_list.append( ["search_term_product_name"] )
target_fields_list.append( ["product_title_product_name"] )
ngrams = [1,2]
aggregation_mode = ["mean", "std", "max", "min", "median"]
for obs_fields, target_fields in zip(obs_fields_list, target_fields_list):
for generator in generators:
for ngram in ngrams:
if ngram == 2:
# since product_name is of length 2, it makes no difference
# for various aggregation as there is only one item
param_list = [ngram, "mean"]
else:
param_list = [ngram, aggregation_mode]
pf = PairwiseFeatureWrapper(generator, dfAll, obs_fields, target_fields, param_list, config.FEAT_DIR, logger)
pf.go()
if __name__ == "__main__":
main(sys.argv[1])
