sedrickkeh / bert-personality-classification

Myers-Briggs Personality Classification and Personality-Specific Language Generation Using Pretrained Language Models

To install BERT, make sure you have pytorch installed, then call

pip install pytorch-pretrained-bert

First run the scraper files to scrape data, then run

python classification/dataset_combiner

Classification:

Run classification model by calling:

python classification/personality_classifier_main.py --do_train --do_eval

Language Generation:

Download pre-trained weights from Google BERT-Base, Cased model (https://storage.googleapis.com/bert_models/2018_10_18/cased_L-12_H-768_A-12.zip) and save to language_generation/data/weights

Convert tensorflow to pytorch-compatible weights by calling:

export BERT_BASE_DIR=language_generation/data/weights/cased_L-12_H-768_A-12

python language_generation/convert_tf_checkpoint_to_pytorch.py \
    --tf_checkpoint_path $BERT_BASE_DIR/bert_model.ckpt \
    --bert_config_file $BERT_BASE_DIR/bert_config.json \
    --pytorch_dump_path $BERT_BASE_DIR/pytorch_model.bin

Create training data by running:

python language_generation/create_training_data.py

Run language generation by calling

python language_generation/personality_language_generation.py \
    --data_dir=language_generation/data/input/<type>
    --output_dir=language_generation/outputs/<type>

For example to run language generation for type ENFJ, run:

python language_generation/personality_language_generation.py \
--data_dir=language_generation/data/input/ENFJ
--output_dir=language_generation/outputs/ENFJ

bert-personality-classification's Issues

Data labelling

I'm interested on using the code for experimenting and label some text reviews that I have. I there any function in code that can I use and predict the personality based on my text input ?