yvonnewu1324/ciphertext-classifier

classify ciphertext using two methods: Pretrained DistilBERT model and Word2Vec + BiLSTM + CNN

Ciphertext Classifier

Implemented two methods to classify ciphertext

• main.ipynb: Pretrained DistilBERT custom embeddings + pretrained DistilBERT. This model has a 88% validation accuracy and 87% on testing data.
• final.ipynb: Word2vec + BiLSTM + CNN and FastText + BiLSTM + CNN. The accuracy hitting around 88% on the dev data.

Dataset

The dataset is split to three categories: train, dev and test. The training and dev data has two columns label and text. Below is the data format:

label	text
0	lkêcê yoúc cêêö y#êjl lw mówám Újám j Úêê# ütlk Úol lkêú z#ê ctöé8ú ówl xoóóú éê#xw#öê#c .
1	yt2ê á#ê6tl lw ê2ê#úwóê x#wö #wÚtócwó 6wüó lw lkê mêú y#té lkjl lktc Úw86 öw2ê üw#mc .

The test data is unlabeled.

Implement Detail

More detail can be find at upload_document.txt.