Next Word Predictor

Predict next word based on last N words typed with two kinds of model

usage

1. N-gram model

• Train and test the model with the following parameters: n, training_file, test_file, output_file, backoff

python3 NGram.py [--N n] [--training_file training_file] [--test_file test_file] [--output_file output_file] [--backoff True/False]

2. LSTM model

• download pretrain Glvoe data

cd data
chmod +x download_glove.sh
./download_glove.sh

• building vocabulary dictionary (data/vocab) with pretrain Glove data(e.g. data/glove.6B.300d.txt) and training corpus(e.g. corpus/train_all.pkl)

python3 vocabulary.py [Glove_file] [corpus_file]

• training Neural Network with training corpus(e.g. corpus/train_all.pkl) and save (data/model_all.pt)
  required: vocabulary file (data/vocab)
  optional: window length

python3 LSTM.py --train corpus_file [--length N]

• testing model(e.g. data/model_all.pt) with testing corpus
  required: vocabulary file (data/vocab)
  optional: window length

python3 LSTM.py --test model_file [--length N]

• Next Word Preditor
  required: vocabulary file (data/vocab), model file(e.g. data/model_all.pt)

python3 app.py