\uc790\uc5f0\uc5b4 \ucc98\ub9ac(Natural Language Processing, NLP)\ub294 \ucef4\ud4e8\ud130\uc640 \uc778\uac04 \uc5b8\uc5b4 \uac04\uc758 \uc0c1\ud638\uc791\uc6a9\uc744 \uac00\ub2a5\ud558\uac8c \ud558\ub294 \uc911\uc694\ud55c \ubd84\uc57c\uc785\ub2c8\ub2e4. \ub525 \ub7ec\ub2dd \uae30\uc220\uc774 \ubc1c\uc804\ud558\uba74\uc11c \uc790\uc5f0\uc5b4 \ucc98\ub9ac \ub610\ud55c \ud070 \ubcc0\ud654\ub97c \uacaa\uace0 \uc788\uc73c\uba70, \uadf8 \uc911 \uc815\uc218 \uc778\ucf54\ub529(Integer Encoding)\uc740 NLP \uc2dc\uc2a4\ud15c\uc5d0\uc11c \ud14d\uc2a4\ud2b8 \ub370\uc774\ud130\ub97c \uc218\uce58\uc801\uc73c\ub85c \ud45c\ud604\ud558\ub294 \ud544\uc218\uc801\uc778 \uacfc\uc815\uc785\ub2c8\ub2e4. \ubcf8 \uac15\uc88c\uc5d0\uc11c\ub294 \uc815\uc218 \uc778\ucf54\ub529\uc758 \uac1c\ub150, \ud544\uc694\uc131, \ubc29\ubc95\ub860, \uadf8\ub9ac\uace0 \uc2e4\uc81c \ub525 \ub7ec\ub2dd \ubaa8\ub378\uc5d0\uc758 \uc801\uc6a9\uc5d0 \ub300\ud574 \uc790\uc138\ud788 \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
\uc815\uc218 \uc778\ucf54\ub529\uc740 \ud14d\uc2a4\ud2b8 \ub370\uc774\ud130\ub97c \uc815\uc218 \ud615\ud0dc\ub85c \ubcc0\ud658\ud558\uc5ec \uba38\uc2e0\ub7ec\ub2dd \ubaa8\ub378\uc774 \uc774\ud574\ud560 \uc218 \uc788\ub3c4\ub85d \ud558\ub294 \uacfc\uc815\uc785\ub2c8\ub2e4. \uc790\uc5f0\uc5b4 \ub370\uc774\ud130\ub294 \ud14d\uc2a4\ud2b8 \ubb38\uc790\uc5f4 \ud615\ud0dc\ub85c \uc874\uc7ac\ud558\uc9c0\ub9cc, \ub300\ubd80\ubd84\uc758 \uba38\uc2e0\ub7ec\ub2dd \uc54c\uace0\ub9ac\uc998\uc740 \uc218\uce58 \ub370\uc774\ud130\ub97c \ucc98\ub9ac\ud558\ub294 \ub370 \ucd5c\uc801\ud654\ub418\uc5b4 \uc788\uc2b5\ub2c8\ub2e4. \ub530\ub77c\uc11c, \ud14d\uc2a4\ud2b8 \ub370\uc774\ud130\uc758 \uc815\uc218 \uc778\ucf54\ub529\uc740 NLP\uc758 \uc804\ucc98\ub9ac \ub2e8\uacc4\uc5d0\uc11c \ub9e4\uc6b0 \uc911\uc694\ud55c \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4.<\/p>\n
\ub300\ubd80\ubd84\uc758 NLP \uc791\uc5c5\uc5d0\uc11c\ub294 \ud14d\uc2a4\ud2b8 \ub370\uc774\ud130\ub97c \uc22b\uc790 \ubca1\ud130 \ud615\ud0dc\ub85c \ubcc0\ud658\ud558\ub294 \uac83\uc774 \ud544\uc218\uc801\uc785\ub2c8\ub2e4. \uba87 \uac00\uc9c0 \uc774\uc720\ub97c \ub4e4\uc5b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4:<\/p>\n
\uc815\uc218 \uc778\ucf54\ub529\uc744 \uc218\ud589\ud558\ub294 \ubc29\ubc95\uc73c\ub85c\ub294 \uc5ec\ub7ec \uac00\uc9c0\uac00 \uc788\uc9c0\ub9cc, \uc77c\ubc18\uc801\uc73c\ub85c \ub2e4\uc74c\uacfc \uac19\uc740 \uacfc\uc815\uc774 \uc788\uc2b5\ub2c8\ub2e4:<\/p>\n
\uc6d0\uc2dc \ud14d\uc2a4\ud2b8 \ub370\uc774\ud130\ub294 \uc815\uc81c(cleaning) \uacfc\uc815\uc744 \ud1b5\ud574 \ubd88\ud544\uc694\ud55c \uae30\ud638, \uc138\ubaa8, \ubaa8\uc9d1\ub2e8\uc5d0\uc11c \uc81c\uac70\ud574\uc57c \ud569\ub2c8\ub2e4. \uc77c\ubc18\uc801\uc778 \ucc98\ub9ac \uacfc\uc815\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n
\uc804\ucc98\ub9ac\ub41c \ud14d\uc2a4\ud2b8\uc5d0\uc11c \uace0\uc720\ud55c \ub2e8\uc5b4\ub97c \ucd94\ucd9c\ud558\uace0 \uac01 \ub2e8\uc5b4\uc5d0 \uace0\uc720\ud55c \uc815\uc218\ub97c \ud560\ub2f9\ud569\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4:<\/p>\n
\ub2e8\uc5b4: [\"\uc0ac\uacfc\", \"\ubc14\ub098\ub098\", \"\ubc30\", \"\uc0ac\uacfc\", \"\uc0ac\uacfc\"]\n\uc815\uc218 \uc778\ucf54\ub529: {\"\uc0ac\uacfc\": 0, \"\ubc14\ub098\ub098\": 1, \"\ubc30\": 2}<\/code><\/pre>\n3. \uc815\uc218 \uc778\ucf54\ub529 \uc801\uc6a9<\/h3>\n
\uac01 \ubb38\uc7a5 \ub0b4\uc758 \ub2e8\uc5b4\ub97c \uace0\uc720\ud55c \uc815\uc218\ub85c \ubcc0\ud658\ud569\ub2c8\ub2e4. \uc608\uc2dc:<\/p>\n
\ubb38\uc7a5: \"\ub098\ub294 \uc0ac\uacfc\ub97c \uc88b\uc544\ud55c\ub2e4.\"\n\uc815\uc218 \uc778\ucf54\ub529: [3, 0, 4, 1]<\/code><\/pre>\n\uc2e4\uc81c \uc0ac\ub840: \ub525 \ub7ec\ub2dd \ubaa8\ub378\uc5d0\uc758 \uc801\uc6a9<\/h2>\n
\uc774\uc81c \uc815\uc218 \uc778\ucf54\ub529\uc758 \uac1c\ub150\uc744 \uc774\ud574\ud588\uc73c\ub2c8, \uc774\ub97c \ub525 \ub7ec\ub2dd \ubaa8\ub378\uc5d0 \uc801\uc6a9\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc608\uc81c\ub85c \uc21c\ud658 \uc2e0\uacbd\ub9dd(RNN)\uc744 \uc0ac\uc6a9\ud558\uc5ec \ud14d\uc2a4\ud2b8 \ubd84\ub958 \ubb38\uc81c\ub97c \ud574\uacb0\ud574\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
1. \ub370\uc774\ud130\uc14b \uc900\ube44<\/h3>\n
\ubb38\uc790 \ub2e8\uc704\ub85c \uc815\uc218 \uc778\ucf54\ub529\ub41c \ub370\uc774\ud130\uc14b\uc744 \uc900\ube44\ud569\ub2c8\ub2e4. \uc608\ub85c IMDB \uc601\ud654 \ub9ac\ubdf0 \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
2. \ubaa8\ub378 \uad6c\ucd95<\/h3>\n
TensorFlow \ub610\ub294 PyTorch\uc640 \uac19\uc740 \ud504\ub808\uc784\uc6cc\ud06c\ub97c \uc0ac\uc6a9\ud558\uc5ec RNN \ubaa8\ub378\uc744 \uad6c\ucd95\ud569\ub2c8\ub2e4:<\/p>\n
import tensorflow as tf\n\nmodel = tf.keras.Sequential([\n tf.keras.layers.Embedding(input_dim=vocab_size, output_dim=64, input_length=max_length),\n tf.keras.layers.SimpleRNN(64),\n tf.keras.layers.Dense(1, activation='sigmoid')\n])<\/code><\/pre>\n3. \ubaa8\ub378 \ud559\uc2b5<\/h3>\n
\ubaa8\ub378\uc744 \ud559\uc2b5\uc2dc\ud0a4\ub294 \uacfc\uc815\uc740 \uc77c\ubc18\uc801\uc778 \ub525 \ub7ec\ub2dd \uc791\uc5c5\uacfc \ub3d9\uc77c\ud569\ub2c8\ub2e4:<\/p>\n
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\nmodel.fit(x_train, y_train, epochs=5, batch_size=32)<\/code><\/pre>\n\uc815\uc218 \uc778\ucf54\ub529\uc758 \uc751\uc6a9 \ubc0f \ud55c\uacc4<\/h2>\n
\uc815\uc218 \uc778\ucf54\ub529\uc740 \uc5ec\ub7ec NLP \uc751\uc6a9 \ud504\ub85c\uadf8\ub7a8\uc5d0\uc11c \uc0ac\uc6a9\ub418\uc9c0\ub9cc, \ud55c\uacc4 \ub610\ud55c \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
1. \ucf54\uc0ac\uc778 \uc720\uc0ac\ub3c4<\/h3>\n
\uc815\uc218 \uc778\ucf54\ub529\uc740 \ub2e8\uc5b4 \uac04\uc758 \uad00\uacc4\ub97c \ubc18\uc601\ud558\uae30 \uc5b4\ub835\uae30 \ub54c\ubb38\uc5d0, \ub2e8\uc5b4\uc758 \uc21c\uc11c\ub098 \uc758\ubbf8\ub97c \uace0\ub824\ud558\uc9c0 \uc54a\uc2b5\ub2c8\ub2e4. \uc774\uac83\uc740 \uc774\ud574\ub3c4\ub97c \ub192\uc774\uae30 \uc704\ud55c \uc790\uc5f0\uc5b4 \ucc98\ub9ac \uc791\uc5c5\uc5d0\uc11c \ubd88\ub9ac\ud55c \uc810\uc774 \ub420 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
2. \uace0\ucc28\uc6d0 \ud76c\uc18c\uc131<\/h3>\n
\ub300\ub7c9\uc758 \uace0\uc720 \ub2e8\uc5b4\uac00 \uc788\uc744 \ub54c, \uadf8\uc5d0 \ub530\ub77c \uc0dd\uc131\ub418\ub294 \uc785\ub825 \ubca1\ud130\ub294 \ub9e4\uc6b0 \ud76c\uc18c(sparse)\ud558\uac8c \ub429\ub2c8\ub2e4. \uc774\ub294 \ubaa8\ub378\uc758 \ud559\uc2b5\uc744 \uc5b4\ub835\uac8c \ud558\uace0, \uc624\ubc84\ud53c\ud305\uc758 \uc704\ud5d8\uc744 \ub192\uc785\ub2c8\ub2e4.<\/p>\n
3. \ub300\uccb4 \uae30\uc220<\/h3>\n
\uc774\ub7ec\ud55c \ud55c\uacc4\ub97c \uadf9\ubcf5\ud558\uae30 \uc704\ud574 Word2Vec, GloVe\uc640 \uac19\uc740 \ub2e8\uc5b4 \uc784\ubca0\ub529 \uae30\ubc95\uc774 \ub3c4\uc785\ub418\uc5c8\uc2b5\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uae30\ubc95\ub4e4\uc740 \ub2e8\uc5b4\ub97c \uace0\ucc28\uc6d0 \ubca1\ud130\ub85c \ubcc0\ud658\ud558\uc5ec \uc758\ubbf8\ub97c \ubcf4\ub2e4 \ud6a8\uacfc\uc801\uc73c\ub85c \ucea1\ucc98\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\uacb0\ub860<\/h2>\n
\uc815\uc218 \uc778\ucf54\ub529\uc740 \ub525 \ub7ec\ub2dd \uae30\ubc18 \uc790\uc5f0\uc5b4 \ucc98\ub9ac\uc5d0\uc11c \ud544\uc218\uc801\uc778 \ub2e8\uacc4\ub85c \uc790\ub9ac\uc7a1\uace0 \uc788\uc2b5\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc744 \ud1b5\ud574 \ud14d\uc2a4\ud2b8\ub97c \uc218\uce58\uc801\uc73c\ub85c \ud45c\ud604\ud568\uc73c\ub85c\uc368 \ubaa8\ub378\uc758 \ud559\uc2b5\uc774 \uac00\ub2a5\ud574\uc9c0\uba70, NLP \ud0dc\uc2a4\ud06c \uc218\ud589\uc5d0 \ud070 \uae30\uc5ec\ub97c \ud558\uace0 \uc788\uc2b5\ub2c8\ub2e4. \ud558\uc9c0\ub9cc \ub2e8\uc5b4 \uac04\uc758 \uad00\uacc4\ub97c \uc81c\ub300\ub85c \ubc18\uc601\ud558\uc9c0 \ubabb\ud558\uace0 \ud76c\uc18c\uc131\uc744 \ucd08\ub798\ud558\ub294 \ub4f1\uc758 \ud55c\uacc4\ub3c4 \uc874\uc7ac\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ub530\ub77c\uc11c \ub2e4\ub978 \uc784\ubca0\ub529 \uae30\ubc95\uacfc \ud568\uaed8 \uc0ac\uc6a9\ud558\uc5ec \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \uadf9\ub300\ud654\ud558\ub294 \uac83\uc774 \ud544\uc694\ud569\ub2c8\ub2e4.<\/p>\n
\ucc38\uace0 \ubb38\ud5cc<\/h2>\n\n- Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to Sequence Learning with Neural Networks. In Advances in Neural Information Processing Systems (NeurIPS).<\/li>\n
- Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed Representations of Words and Phrases and their Compositionality. In Advances in Neural Information Processing Systems (NeurIPS).<\/li>\n
- Bahdanau, D., Cho, K., & Bengio, Y. (2014). Neural Machine Translation by Jointly Learning to Align and Translate. In International Conference on Learning Representations (ICLR).<\/li>\n<\/ul>\n
<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"