-
Notifications
You must be signed in to change notification settings - Fork 0
/
Ngram_model.py
126 lines (103 loc) · 4.14 KB
/
Ngram_model.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
import json
import os
from nltk import trigrams
from collections import defaultdict
import random
import settings
import pandas as pd
def create_trigram_model():
path = settings.ngram_dir
if not os.path.isdir(path):
os.mkdir(path)
if os.path.exists(settings.ngram_model):
return
print('----- Analyzing Trigram Model ------')
songs_list_as_tuples = []
df = pd.read_csv(settings.csv_file_name, encoding="utf-8-sig")
df = df[df['lyrics'].notna()]
# df = df[(df['year'] >= 80) & (df['year'] <= 80)]
# Convert lyrics to a list of words, while '\n' is a word too
for index, row in df.iterrows():
song_lines_list = row['lyrics'].split('\n')
for i, line in enumerate(song_lines_list):
song_lines_list[i] = line + ' \n'
song_lines_list = ' '.join(song_lines_list)
song_lines_list = song_lines_list.split(' ')
song_as_words = list(filter(''.__ne__, song_lines_list))
song_as_words = (song_as_words,row['weeks'])
songs_list_as_tuples.append(song_as_words)
# Create a placeholder for model
model = defaultdict(lambda: defaultdict(lambda: 0))
# Count frequency of co-occurance
for song in songs_list_as_tuples:
for w1, w2, w3 in trigrams(['~~~','~~~']+song[0]+['^^^','^^^']):
model[w1+'@'+w2][w3] += song[1]
# Let's transform the counts to probabilities
for key in model:
total_count = float(sum(model[key].values()))
for w3 in model[key]:
model[key][w3] /= total_count
print('----- Storing Data ------')
with open(settings.ngram_model, 'w') as json_file:
json.dump(model, json_file, indent=4)
json_file.seek(0)
print('----- DONE ------')
def generate_song(model):
text_temp = ['~~~','~~~']
text = ['~~~','~~~']
sentence_finished = False
num_of_words = 700
while not sentence_finished:
# select a random probability threshold
if len(text_temp) > num_of_words:
break
r = random.random()
accumulator = .0
for word in model['@'.join(text_temp[-2:])].keys():
accumulator += model['@'.join(text_temp[-2:])][word]
# select words that are above the probability threshold
if accumulator >= r:
counter = word.count('%')
if counter > 0:
pres = '%'*counter
start = int('1'+('0'*(counter-1)))
end = int('9'+('0'*(counter-1)))
new_word = word.replace(pres, str(random.randint(start, end)))
text.append(new_word)
text_temp.append(word)
else:
text.append(word)
text_temp.append(word)
break
if text_temp[-2:] == ['^^^', '^^^']:
sentence_finished = True
# Generate a song as list of words & handle '/n'
song_as_words = [t for t in text if t]
for i in range (0, len(song_as_words)-1):
if song_as_words[i]=='\n':
song_as_words[i+1]='\n'+song_as_words[i+1]
song_as_words = list(filter('\n'.__ne__, song_as_words))
for i in range(0, len(song_as_words) - 1):
if song_as_words[i] == '\n\n':
song_as_words[i + 1] = '\n\n' + song_as_words[i + 1]
song_as_words = list(filter('\n\n'.__ne__, song_as_words))
song = ' '.join(song_as_words)
song = song[8:-8]
return song
# Generate song with fix statistics
# # Calculate Statistics
# with open(settings.statistics_json, 'r') as json_file:
# statistics = json.load(json_file)
# json_file.seek(0)
#
# words_per_line = statistics['average_words_per_line']
# lines_per_stanza = statistics['average_lines_per_stanza']
#
# # split lines every 'words_per_line' number
# song = song.split()
# song = '\n'.join([' '.join(song[i:i + words_per_line]) for i in range(0, len(song), words_per_line)])
# # split lines to stanzas
# song = song.split('\n')
# song = '\n\n'.join(['\n'.join(song[i:i + lines_per_stanza]) for i in range(0, len(song), lines_per_stanza)])
# # return song
# return song