Hide keyboard shortcuts

Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

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

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

import pathlib 

import sqlite3 

import logging 

import datetime 

import tempfile 

from xml.etree import ElementTree 

 

import flask 

import cophi 

import numpy as np 

from werkzeug.utils import secure_filename 

 

TEMPDIR = tempfile.gettempdir() 

DATABASE_URI = str(pathlib.Path(TEMPDIR, "topicsexplorer.db")) 

LOGFILE = str(pathlib.Path(TEMPDIR, "topicsexplorer.log")) 

 

def init_logging(level=logging.DEBUG): 

# Set up basic configuration: 

logging.basicConfig(level=level, 

format="%(message)s", 

filename=LOGFILE, 

filemode="w") 

# Ignore messages from flask and werkzeug: 

logging.getLogger("flask").setLevel(logging.ERROR) 

logging.getLogger("werkzeug").setLevel(logging.ERROR) 

 

 

def get_status(): 

# Open logfile: 

with pathlib.Path(LOGFILE).open("r", encoding="utf-8") as logfile: 

# Read lines: 

messages = logfile.readlines() 

# Select and strip the last one: 

message = messages[-1].strip() 

# Format the log message: 

message = format_logging(message) 

# Get current time: 

now = datetime.datetime.now().strftime("%H:%M:%S") 

return f"{now}<br>{message}" 

 

def format_logging(message): 

if "n_documents" in message: 

n = message.split("n_documents: ")[1] 

return f"Number of documents: {n}." 

elif "vocab_size" in message: 

n = message.split("vocab_size: ")[1] 

return f"Number of types: {n}." 

elif "n_words" in message: 

n = message.split("n_words: ")[1] 

return f"Number of tokens: {n}." 

elif "n_topics" in message: 

n = message.split("n_topics: ")[1] 

return f"Number of topics: {n}." 

elif "n_iter" in message: 

# Logging might freeze at this point, 

# so better log something like this: 

return "Initializing topic model." 

elif "log likelihood" in message: 

iteration, log_likelihood = message.split("> log likelihood: ") 

return f"Iteration {iteration[1:]}, log-likelihood: {log_likelihood}." 

else: 

return message 

 

 

def get_db(): 

"""Establish connection to database. 

""" 

if "db" not in flask.g: 

# Establish connection to pointed URI: 

flask.g.db = sqlite3.connect(DATABASE_URI) 

return flask.g.db 

 

 

def close_db(e=None): 

"""Close connection to database. 

""" 

# Check if a connection was created: 

db = flask.g.pop("db", None) 

if db is not None: 

# Close connection if exists: 

db.close() 

 

 

def init_db(app): 

"""Initialize database and create tables. 

""" 

db = get_db() 

with app.open_resource("schema.sql") as schemafile: 

schema = schemafile.read().decode("utf-8") 

db.executescript(schema) 

db.commit() 

close_db() 

 

 

def get_data(corpus, topics, iterations, stopwords, mfw): 

"""Get input data. 

""" 

# Get text files, number of topics and number of iterations: 

data = {"corpus": flask.request.files.getlist("corpus"), 

"topics": int(flask.request.form["topics"]), 

"iterations": int(flask.request.form["iterations"])} 

# Get stopword list, if user selected one: 

if flask.request.files.get("stopwords", None): 

data["stopwords"] = flask.request.files["stopwords"] 

# Use most frequent words threshold otherwise: 

else: 

data["mfw"] = int(flask.request.form["mfw"]) 

return data 

 

 

def insert_into_textfiles(values): 

"""Insert text files into table. 

""" 

# Connect to database: 

db = get_db() 

# Insert values into table: 

for textfile in values: 

# Get title and text: 

title, text = load_textfile(textfile) 

# Execute SQL: 

db.execute(""" 

INSERT INTO textfiles (title, text)  

VALUES(?, ?); 

""", 

[title, text]) 

db.commit() 

close_db() 

 

 

def select_textfiles(): 

cursor = get_db().cursor() 

return cursor.execute(""" 

SELECT title, text  

FROM textfiles; 

""") 

 

 

def get_documents(textfiles): 

for textfile in textfiles: 

title, text = textfile 

yield cophi.model.Document(text, title) 

 

def load_textfile(textfile): 

"""Load textfile and get title. 

""" 

# Get filename: 

filename = pathlib.Path(secure_filename(textfile.filename)) 

# Get title: 

title = filename.stem 

# Get file extension: 

suffix = filename.suffix 

# Read file: 

text = textfile.read().decode("utf-8") 

# If suffix implies any markup, remove it: 

if suffix in {".xml", ".html"}: 

text = remove_markup(text) 

return title, text 

 

 

def remove_markup(text): 

"""Parse string and remove markup. 

""" 

tree = ElementTree.fromstring(text) 

plaintext = ElementTree.tostring(tree, 

encoding="utf8", 

method="text") 

return plaintext.decode("utf-8") 

 

def get_stopwords(data, corpus): 

if "stopwords" in data: 

_, stopwords = load_textfile(data["stopwords"]).split("\n") 

else: 

stopwords = corpus.mfw(data["mfw"]) 

return stopwords 

 

 

def preprocess(data): 

# Query text files: 

textfiles = select_textfiles() 

# Get cophi.model.Document object: 

documents = get_documents(textfiles) 

# Create cophi.model.Corpus object: 

corpus = cophi.model.Corpus(documents) 

# Get stopwords: 

stopwords = get_stopwords(data, corpus) 

# Get hapax legomena: 

hapax = corpus.hapax 

# Join both lists: 

features = set(stopwords).union(set(hapax)) 

# Clean document-term matrix: 

dtm = corpus.drop(corpus.dtm, features) 

# Get sizes: 

sizes = corpus.num_tokens 

# Convert to a NumPy array and return: 

return dtm.values, dtm.columns, dtm.index, sizes.values 

 

def get_topics(model, vocabulary, maximum=100): 

for i, distribution in enumerate(model.topic_word_): 

yield np.array(vocabulary)[np.argsort(distribution)][:-maximum-1:-1] 

 

 

def get_similarities(matrix): 

d = matrix.T @ matrix 

norm = (matrix * matrix).sum(0, keepdims=True) ** .5 

return d / norm / norm.T 

 

def normalize(matrix, sizes): 

# Multiply weights with word frequencies: 

matrix = matrix * sizes.sum() 

# Normalize with text lengths: 

return matrix / sizes 

 

 

def scale(vector): 

return np.interp(vector, (vector.min(), vector.max()), (40, 100))