自社のサービスで、テキストを分類したりラベリングしたりということをやりたくなったので、文章タスクが得意らしいRNN(LSTM)で単純な分類問題を解いてみたい。
スパムメッセージ分類がちょうど良い難易度っぽいので、以下エントリのコードをお借りして動くところまでやってみた。
https://ohke.hateblo.jp/entry/2019/04/27/154500
環境構築
大体この手のはanaconda使えと書いてあるが、ぼくはなんとなく使わない派(それほど深い理由はない)なので、pypenv+virtualenvを使います。コードの実行は定番のjupyter notebookを使います。
$ mkdir keras-rnn-spam-filter $ cd keras-rnn-spam-filter
今回のプロジェクト専用の仮想環境を作る。
$ pyenv virtualenv 3.7.3 keras-rnn-spam-filter $ pyenv local keras-rnn-spam-filter $ pyenv versions system 3.6.9 3.7.3 * keras-rnn-spam-filter (set by /Users/atsushiharada/source/keras-rnn-spam-filter/.python-version)
あれ notebook がインストールできないやんけ。opensslの問題ぽい。
$ pip install notebook ✘ 1
pip is configured with locations that require TLS/SSL, however the ssl module in Python is not available.
Collecting jupyter
Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by 'SSLError("Can't connect to HTTPS URL because the SSL module is not available.")': /simple/jupyter/
Retrying (Retry(total=3, connect=None, read=None, redirect=None, status=None)) after connection broken by 'SSLError("Can't connect to HTTPS URL because the SSL module is not available.")': /simple/jupyter/
Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'SSLError("Can't connect to HTTPS URL because the SSL module is not available.")': /simple/jupyter/
Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'SSLError("Can't connect to HTTPS URL because the SSL module is not available.")': /simple/jupyter/
Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'SSLError("Can't connect to HTTPS URL because the SSL module is not available.")': /simple/jupyter/
Could not fetch URL https://pypi.org/simple/jupyter/: There was a problem confirming the ssl certificate: HTTPSConnectionPool(host='pypi.org', port=443): Max retries exceeded with url: /simple/jupyter/ (Caused by SSLError("Can't connect to HTTPS URL because the SSL module is not available.")) - skipping
Could not find a version that satisfies the requirement jupyter (from versions: )
No matching distribution found for jupyter
pip is configured with locations that require TLS/SSL, however the ssl module in Python is not available.
Could not fetch URL https://pypi.org/simple/pip/: There was a problem confirming the ssl certificate: HTTPSConnectionPool(host='pypi.org', port=443): Max retries exceeded with url: /simple/pip/ (Caused by SSLError("Can't connect to HTTPS URL because the SSL module is not available.")) - skipping
ここを参考に再インストールしてみるもダメだった。
https://qiita.com/drumehiron/items/da79d837ffea3723aafe
$ brew unlink openssl $ brew reinstall openssl
reinstall時に表示されるログの通り、パスを通してみたがこれも解決せず。
$ echo 'export PATH="/usr/local/opt/openssl@1.1/bin:$PATH"' >> ~/.zshrc $ source ~/.zshrc
たぶんpythonインストール時にopensslが使用されるっぽいんだけど、それが以前のバージョンと異なっているからエラーになっている？
pythonのインストールからやり直してみる。
$ pyenv uninstall 3.7.3 $ pyenv install 3.7.3
再び仮想環境を作る。
$ pyenv virtualenv 3.7.3 keras-rnn-spam-filter $ pyenv local keras-rnn-spam-filter $ pyenv versions system 3.6.9 3.7.3 3.7.3/envs/keras-rnn-spam-filter * keras-rnn-spam-filter (set by /Users/atsushiharada/source/keras-rnn-spam-filter/.python-version)
再び notebook をインストール。今回は上手く行ったぞ。
$ pip install notebook
起動。
$ jupyter notebook
スパム分類器の実装
必要なライブラリをインストールする。
$ pip install pandas scikit-learn keras tensorflow
こちらからスパム分類のデータセットをダウンロードしてくる。
https://archive.ics.uci.edu/ml/datasets/sms+spam+collection
解凍後のSMSSpamCollection にこんな感じのデータが入ってる。
ham Go until jurong point, crazy.. Available only in bugis n great world la e buffet... Cine there got amore wat... ham Ok lar... Joking wif u oni... spam Free entry in 2 a wkly comp to win FA Cup final tkts 21st May 2005. Text FA to 87121 to receive entry question(std txt rate)T&C's apply 08452810075over18's ham U dun say so early hor... U c already then say... ham Nah I don't think he goes to usf, he lives around here though spam FreeMsg Hey there darling it's been 3 week's now and no word back! I'd like some fun you up for it still? Tb ok! XxX std chgs to send, £1.50 to rcv ham Even my brother is not like to speak with me. They treat me like aids patent. ham As per your request 'Melle Melle (Oru Minnaminunginte Nurungu Vettam)' has been set as your callertune for all Callers. Press *9 to copy your friends Callertune spam WINNER!! As a valued network customer you have been selected to receivea £900 prize reward! To claim call 09061701461. Claim code KL341. Valid 12 hours only. spam Had your mobile 11 months or more? U R entitled to Update to the latest colour mobiles with camera for Free! Call The Mobile Update Co FREE on 08002986030 ...
データセットをプロジェクト配下に移動。
$ mv ~/Downloads/smsspamcollection/SMSSpamCollection .
参考にしたエントリのコードをそのままコピペして動かしてみる。
from sklearn.model_selection import train_test_split
X_train, X_test, Y_train, Y_test = train_test_split(
dataset_df[['text']], dataset_df[['category']],
test_size=0.2, random_state=0
)
print(X_train.shape, X_test.shape, Y_train.shape, Y_test.shape)
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
max_len = 100
tokenizer = Tokenizer()
tokenizer.fit_on_texts(X_train['text'])
x_train = tokenizer.texts_to_sequences(X_train['text'])
x_test = tokenizer.texts_to_sequences(X_test['text'])
for text, vector in zip(X_train['text'].head(3), x_train[0:3]):
print(text)
print(vector)
x_train = pad_sequences(x_train, maxlen=max_len) x_test = pad_sequences(x_test, maxlen=max_len) print(x_train[0])
rom keras.models import Sequential from keras.layers import LSTM, Dense, Embedding vocabulary_size = len(tokenizer.word_index) + 1 model = Sequential() model.add(Embedding(input_dim=vocabulary_size, output_dim=32)) model.add(LSTM(16, return_sequences=False)) model.add(Dense(1, activation='sigmoid')) model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy']) model.summary()
y_train = Y_train['category'].values
y_test = Y_test['category'].values
history = model.fit(
x_train, y_train, batch_size=32, epochs=10,
validation_data=(x_test, y_test)
)
学習出来た。
スパムっぽい文章をいれて分類させてみる
データセットのスパムメッセージ文面を見ながら作ってみたが、自分で英語のスパムメッセージ作るってなかなか難しいのな。。。
こんなGoogle翻訳を使ってメッセージ作ってみた。我ながらしょうもない文面を書いてしまった自覚があるので、日本語にはしない方がいいぞ。
Are you single? Use this site to find women in their 20s who want to have sex right away! Pay $ 100 Now and Get Free! 08717890890
データセットを見ていると、spamラベルがついているメッセージには怪しげなURLがついてることが多いので、怪しげなURL(当然いい加減なURL)を付け加えた版。
Are you single? Use this site to find women in their 20s who want to have sex right away! Pay $ 100 Now and Get FREE! 08717890890 http://porn.sex/xxx.asp?o=13543
あとは普通のお花見に誘うメッセージ。
Hello. It is nice weather. It's time to see the cherry blossoms, so how about cherry blossom viewing together?
texts = [
"Are you single? Use this site to find women in their 20s who want to have sex right away! Pay $ 100 Now and Get FREE! 08717890890",
"Are you single? Use this site to find women in their 20s who want to have sex right away! Pay $ 100 Now and Get FREE! 08717890890 http://porn.sex/xxx.asp?o=13543",
"Hello. It is nice weather. It's time to see the cherry blossoms, so how about cherry blossom viewing together"
]
x = tokenizer.texts_to_sequences(texts)
x = pad_sequences(x, maxlen=max_len)
y_pred = model.predict_classes(x)
print(y_pred)
結果
URLがついていない方はスパム判定してくれませんでした。学習モデルが悪いというより、ぼくのスパムメッセージ作成スキルが足りないせいっぽい。お花見に誘うメッセージはあやしいところはないので当然非スパムと判定された。
RNN(LSTM)でシンプルな分類問題を解くやり方は分かったので、次何しよっかな。
