def train_tf_idf(file_list, **kwargs):
# Default params
tf_idf_params = {
'input': 'filename',
'encoding': 'utf-8',
'decode_error': 'replace',
'strip_accents': 'unicode',
'lowercase': True,
'analyzer': 'word',
'stop_words': 'english',
'token_pattern': r'(?u)\b[a-zA-Z_][a-zA-Z0-9_]+\b',
'ngram_range': (1, 2),
'max_features': 5000,
'norm': 'l2',
'use_idf': True,
'smooth_idf': True,
'sublinear_tf': True,
'max_df': 1.0,
'min_df': 1}
# Update with kwargs if any
tf_idf_params.update(kwargs)
train_list = list(file_list) # creates copy
shuffle(train_list)
# compute tfidf vectors with scikits
vectorizer = TfidfVectorizer(**tf_idf_params)
vectorizer.fit(train_list)
tf_idf_matrix = vectorizer.fit_transform(file_list)
# CHange input type to content (string) dfor later
vectorizer.input = 'content'
return vectorizer, tf_idf_matrix
