A Minecraft game plugin that can generate different terrains according to user descriptions, combining PCG algorithm and NLP.
The idea behind this project is the creation of a game world in Minecraft that functions like a black box. It is by definition created by a seed, which is an integer number. For example, "seed 42". However, this process brings uncertainty and misunderstanding of how the world actually is, as people can not imagine a world just with a number, even for hardcore players. Therefore, a plugin can take in the descriptive sentences input by players and then generate a game world accordingly. This will make the entire process more understandable and transparent.
This Python project, developed by the team members of the MGAA course at LIACS in 2023 and further refined by R. Ma, utilizes a variety of libraries including Numpy, Spacy, PyTorch, and GDPC to perform natural language processing and terrain generation in a simulated environment. The code is designed to tokenize English text, build vocabularies from datasets, transform sequences, and classify them using a RNN (Recurrent Neural Network). It also interfaces with the Minecraft game to create terrain based on the classifications made by the RNN.
1.0
- English Text Tokenization: Tokenizes English sentences using Spacy's NLP model.
- Vocabulary Building: Constructs vocabularies from given text data.
- Sequence Transformation: Applies transformations to sequences of tokens.
- RNN Classification: Uses a RNN model for classifying sequences into different categories.
- Terrain Generation: Interfaces with Minecraft to generate terrain based on classification results.
- Plant and Tree Planting: Depending on the classified terrain type, different plants and trees are planted.
- Python 3.x
- PyTorch
- TorchText
- Spacy
- GDPC (with Minecraft and GDMC HTTP interface)
- NumPy
- Scikit-Learn
- tqdm
- A working Minecraft setup with the GDMC HTTP mod installed.
- Clone the repository or download the source code.
- Install required Python packages:
pip install numpy spacy torch torchtext sklearn tqdm gdpc - Install Spacy's English model:
python -m spacy download en_core_web_sm
- Run the script using Python.
- Start running the Minecraft game and open a super flat world.
- Input a description of the world when prompted.
- The script will interact with Minecraft to generate terrain based on the classification.
For further inquiries, please contact the repo owner.
This project is part of an academic course and is intended for educational purposes only. It requires a specific setup with Minecraft and GDPC HTTP interface for full functionality.