hiimbach / sentiment_analysis

Sentiment Analysis

PhoBERT + 2 layer BiLSTM for Sentiment Analysis 🙆‍♀️

1. Download data from this link. Extract data and set value of SOURCE_FOLDER in train_lstm_phobert.py to path of dataset.

2. Download pretrained model here. Set value of state_dict_path to the path of pretrained_model.

3. Remember to run pip3 install -r requirements.txt to install all nescessary packages.

4. To install CocCoc Tokenizer, please run:

    git clone https://github.com/coccoc/coccoc-tokenizer.git

When coccoc-tokenizer is already cloned. Run these commands on terminal to install it.

$ git clone https://github.com/coccoc/coccoc-tokenizer
$ cd coccoc-tokenizer && mkdir build && cd build
$ cmake -DBUILD_PYTHON=1 -DCMAKE_INSTALL_PREFIX=$HOME/.local ..
$ make install

5. Now you can train your own model, by running:

    python3 train_lstm_phobert.py

You can also run to evaluate pretrained model by running:

    python3 evaluate.py

If you want to test pretrained model on your data. Please put your data into an txt file. You can see an example at test.txt. After that, you can run:

    python3 test_lstm_phobert.py