def remove_common_words(self, words):
"""
Removes a word or list of words that are considered common from all self.docs collection
:param words: string or list of strings
:return:
"""
# if words is a single word convet it into a list for versatility of application of next lines of code
if isinstance(words, str): words = [words]
# if words are not stemmed we want to also remove their stemmed version from self.docs, so add them to words
"""
stemmed_words = [tm.stem(word) for word in words]
words = words.extend(word for word in stemmed_words if word not in words)
"""
words.extend(
tm.stem(word) for word in words if tm.stem(word) not in words)
# remove words from self.docs
self.docs = [
filter(lambda word: word not in words, doc) for doc in self.docs
]
def __init__(self, controlled_vocab_file):
self.terms = {}
with open(controlled_vocab_file, 'r') as infile:
for num, line in enumerate(infile):
self.terms[line.splitlines()[0]] = []
self.terms_flat = [stem(term) for phrase in self.terms for term in phrase.split()]
# maps index_pos to _id
self.doc_ids = []
self.tdm = TermDocumentMatrix()
# lookup table of index positions for terms on TDM
self.term_index_lookup = {}
# nd array structure for TDM
self.matrix = np.array([])
# output dict; comp site for phrase matching
self.out_dict = {}
# pseudo hash trick dict
self.term_index_dict = {}
def compute_phrase_matches(self):
"""computes phrase matches as dot products of term vectors
"""
for search_phrase in self.terms.keys():
terms = [stem(term) for term in search_phrase.split()]
# retrieve column vector for unigrams
if len(terms) == 1:
try:
# grab column for term and skip first row
self.out_dict[search_phrase] = self.matrix[:, dict[search_phrase]][1:]
except (KeyError, TypeError):
pass
# get vectors to multiply; compute dot product of vectors
else:
try:
dot_product = reduce(operator.__mul__, [self.matrix[:, self.term_index_dict[a]][1:].astype(int) for a in terms], 1)
self.out_dict[search_phrase] = dot_product
except (KeyError, TypeError):
pass
tags = {}
for search_phrase in self.out_dict:
# maintain original doc index pos with enumeration
reduced = [a for a in enumerate(self.out_dict[search_phrase]) if a[1] != 0 and a[1] != '0']
for pair in reduced:
tags.setdefault(pair[0], [])
# doc_index_pos: tagged_phrase
tags[pair[0]].append(search_phrase)
doc_lookup = {}
for _id in enumerate(self.doc_ids):
doc_lookup[_id[0]] = _id[1]
out = {}
for search_phrase in tags:
out[doc_lookup[search_phrase]] = tags[search_phrase]
# doc id and list of phrase matches;
matches = [[i[0], ','.join(i[1])] for i in out.iteritems()]
return matches
