NLP In Video Games

This repository contains the implementation of NLP In Video games created in order to test its functionality.

The repository contains the following:

• Test Game
• Joint Model for intent classification and slots filling
• Question Answering model used to answer questions

Test Game

The test game is a simple simulation in which several locations were created. Those locations are used in order to test the NLP models.

The game contains a simple chat system used to communicate with the BOT using Natural Language.

In order to run the game Unreal Engine 4.25 is required. The game can run either from the engine itself or can be built to test it.

Joint Model

Joint model is a model which, from a given text, finds out what the intent of the user is and what the slots in the input text are.

For more information about the Joint Model see JointModel directory.

Question Answering

Question Answering is a model which answers to the given question sent as a text.

For more information about the Question Answering Model see QuestionAnswering directory.

Examples

Question Answering

1. Where can i heal?

2. which is th hardest dungeon in th game?

3. doom dungeon required level

Follow Action

1. Get me to teleportr

2. Get me to arm shp

3. show me castle of doom entrance

References

• Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova, "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding", October 2018
• Qian Chen, Zhu Zhuo, Wen Wang, "BERT for Joint Intent Classification and Slot Filling", February 2019
• OpenAI Team, "Language Models are Few-Shot Learners", May 2020
• "joint-intent-classification-and-slot-filling-based-on-BERT"
• Moscow Institute of Physics and Technology (MIPT), "DeepPavlov"