This repo is 끝말잇기 ai.

• Python

This repo contain fully working code. you can train this ai with custom dataset. and code for evaluate it. and also you can use this ai inference.

If you train this model, This model will continue to iterate over and over again these steps.

1. Will play game by its ownself until the end.
2. When the game is over. evaluating words used in this game. and save its result.

A trained model has a couple of data that winning rate, frequency,...etc representative that word. inference step is a combine this data and make choice according to its score.

1. Prepare dataset. (this repo contain sample dataset. you can pass this step)

Make your dataset. and add them data/words.txt. training data must fallow these template.

가돌리늄
가듁
가래톳
가렌
가렛좃

2. Training WOM model

python index.py

Variable episodes is the number of times to repeat. You can adjust this. If train finished, trained_model.pickle file will be created.

3. inference