Push training code for text embeddings using Fasttext, same as speech…

…2vec paper

more_experiments/train_embeddings_helpers.py

@@ -0,0 +1,65 @@
+import os
+import pandas as pd
+import spacy
+
+def get_alignment_filenames(path_to_alignments_folder):
+    '''Get all file names in the annotated word alignments'''
+    root_path = 'data/LibriSpeech-Alignments/LibriSpeech_Text_Alignments/'
+    folders = ['dev-clean', 'dev-other', 'test-clean', 'test-other', 'train-clean-100', 'train-clean-360', 'train-other-500']
+    filenames = dict()
+    for folder in folders:
+        grab_files = []
+        for path, subdirs, files in os.walk(root_path + folder):
+            for name in files:
+                grab_files.append(os.path.join(path, name))
+        filenames[folder] = grab_files
+    return filenames # Returns a dictionarty with {subfolder_name:[filenames_in_subfolder]}
+
+def get_text_from_files(filenames_dict):
+    '''Will create a dataframe per folder, with all the text combined as the column
+    Questions:
+    Do we need to keep silence?
+    '''
+    all_folders_df = pd.DataFrame()
+    for folder, filenames in filenames_dict.items():
+        folder_df = pd.DataFrame()
+        for filename in filenames:
+            if filename[-3:] == 'txt':
+                file_text = extract_text_single_file(filename)
+                folder_text = {'Folder':[folder], 'Filename':[filename], 'Text':[file_text]}
+                text_df = pd.DataFrame(folder_text)
+                folder_df = folder_df.append(text_df)
+        all_folders_df = all_folders_df.append(folder_df)
+    return all_folders_df
+
+def extract_text_single_file(file_path):
+    with open(file_path, 'r') as f:
+        content = f.read().splitlines()
+    all_text = ''
+    for line in content:
+        text = line.split()[1].split(',')
+        text = [word for word in text if word.isalnum()]
+        all_text += ' '.join(text) + ' '
+    return all_text
+
+def main():
+    filenames_dict = get_alignment_filenames('data/LibriSpeech-Alignments/LibriSpeech_Text_Alignments')
+    df = get_text_from_files(filenames_dict)
+    return df
+
+class TextPreprocess():
+    def __init__(self):
+        self.nlp = spacy.load('en')
+
+    def lemmatize(self, text_chunk):
+        text_model = self.nlp(text_chunk)
+        new_chunk = ' '.join([word.lemma_ if word.lemma_ != '-PRON-' else word.text for word in text_model])
+        return new_chunk
+
+    def remove_stopwords(self, text_chunk):
+        text_model = self.nlp(text_chunk)
+        new_chunk = ' '.join([word.text for word in text_model if not word.is_stop ])
+        return new_chunk
+
+if __name__ == "__main__":
+    main()