From 6622458fe790924d142fa29b15991f10583847b9 Mon Sep 17 00:00:00 2001
From: Nirant K <nirant@soroco.com>
Date: Thu, 15 Nov 2018 15:03:12 +0000
Subject: [PATCH] Add aclImdb extractor

---
 fastai_contrib/utils.py | 85 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 84 insertions(+), 1 deletion(-)

diff --git a/fastai_contrib/utils.py b/fastai_contrib/utils.py
index 5f40940..f92fff6 100644
--- a/fastai_contrib/utils.py
+++ b/fastai_contrib/utils.py
@@ -3,11 +3,15 @@ Utility methods for data processing.
 """
 import pandas as pd
 import numpy as np
+import fire
 from fastai import F, to_device
 import torch
 from tqdm import tqdm
 import re
 import csv
+import pathlib
+import tarfile
+from sklearn import model_selection
 
 EOS = '<eos>'
 UNK = '<unk>'
@@ -17,6 +21,81 @@ PAD_TOKEN_ID = 1
 number_match_re = re.compile(r'^([0-9]+[,.]?)+$')
 number_split_re = re.compile(r'([,.])')
 
+CLASSES = ['neg', 'pos', 'unsup']
+
+
+def get_texts(path):
+    texts, labels = [],[]
+    for idx, label in enumerate(CLASSES):
+        for fname in (path/label).glob('*.*'):
+            texts.append(fname.open('r', encoding='utf-8').read())
+            labels.append(idx)
+    return np.array(texts), np.array(labels)
+
+
+def prepare_imdb(file_path: str, prepare_lm = False):
+    """
+    function to extract aclImdb and combine into fastai standard format of labels and then text
+    columns
+
+    Args:
+        file_path: path to the aclImdb.tgz
+        prepare_lm (bool): prepare file for language model finetuning
+
+    Returns:
+        None
+    """
+
+    file_path = pathlib.Path(file_path)
+    dir_path = pathlib.Path(file_path.stem).resolve()
+    assert tarfile.is_tarfile(file_path), "this is not a valid targz file"
+
+    if not dir_path.exists():
+        print(f"Extracting {file_path} to {dir_path}. This may take a long time...")
+        tgz_file = tarfile.open(file_path)
+        tgz_file.extractall()
+        assert dir_path.exists()
+        print(f"Extracted to {dir_path}")
+
+    CLAS_PATH = dir_path / 'imdb_clas'
+    CLAS_PATH.mkdir(exist_ok=True)
+
+    LM_PATH = dir_path /'imdb_lm'
+    LM_PATH.mkdir(exist_ok=True)
+
+    # processing the split files to create train.csv and test.csv in fastai format
+    col_names = ['labels', 'text']
+    trn_texts, trn_labels = get_texts(dir_path/ 'train')
+    val_texts, val_labels = get_texts(dir_path / 'test')
+    np.random.seed(42)
+    trn_idx = np.random.permutation(len(trn_texts))
+    val_idx = np.random.permutation(len(val_texts))
+    trn_texts = trn_texts[trn_idx]
+    val_texts = val_texts[val_idx]
+    trn_labels = trn_labels[trn_idx]
+    val_labels = val_labels[val_idx]
+
+    df_trn = pd.DataFrame({'text': trn_texts, 'labels': trn_labels}, columns=col_names)
+    df_val = pd.DataFrame({'text': val_texts, 'labels': val_labels}, columns=col_names)
+    print(f"df_trn has {len(df_trn)} rows, while df_val has {len(df_val)} rows")
+    print(f"Writing them to {CLAS_PATH}")
+    df_trn[df_trn['labels'] != 2].to_csv(CLAS_PATH / 'train.csv', header=False, index=False)
+    df_val.to_csv(CLAS_PATH / 'test.csv', header=False, index=False)
+
+    (CLAS_PATH / 'classes.txt').open('w', encoding='utf-8').writelines(f'{o}\n' for o in CLASSES)
+
+    if prepare_lm:
+        print("Preparing LM data")
+        trn_texts, val_texts = model_selection.train_test_split(
+            np.concatenate([trn_texts, val_texts]), test_size=0.1)
+        print(f"trn_texts has {len(trn_texts)} samples, while val_texts has {len(val_texts)} rows")
+        print(f"Writing them to {LM_PATH}")
+        df_trn = pd.DataFrame({'text': trn_texts, 'labels': [0] * len(trn_texts)}, columns=col_names)
+        df_val = pd.DataFrame({'text': val_texts, 'labels': [0] * len(val_texts)}, columns=col_names)
+
+        df_trn.to_csv(LM_PATH / 'train.csv', header=False, index=False)
+        df_val.to_csv(LM_PATH / 'test.csv', header=False, index=False)
+
 
 def replace_number(token):
     """Replaces a number and returns a list of one or multiple tokens."""
@@ -124,4 +203,8 @@ class TextReader():
     def get_batch(self, i, seq_len):
         source = self.data
         seq_len = min(seq_len, len(source) - 1 - i)
-        return source[i:i+seq_len], source[i+1:i+1+seq_len].view(-1)
\ No newline at end of file
+        return source[i:i+seq_len], source[i+1:i+1+seq_len].view(-1)
+
+
+if __name__ == "__main__":
+    fire.Fire()  # allows using all functions via CLI e.g. python utils.py prepare_imdb aclImdb.tgz
\ No newline at end of file