Skip to content

Latest commit

 

History

History
96 lines (74 loc) · 9.88 KB

README.md

File metadata and controls

96 lines (74 loc) · 9.88 KB
Raw

MindCraft

Authors: Cristian-Paul Bara*, Sky CH-Wang*, Joyce Chai

This is the official code repository for the paper (anthology link):

Cristian-Paul Bara, Sky CH-Wang, and Joyce Chai. 2021. MindCraft: Theory of Mind Modeling for Situated Dialogue in Collaborative Tasks. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP).

@inproceedings{bara-etal-2021-mindcraft,
    title = "{M}ind{C}raft: Theory of Mind Modeling for Situated Dialogue in Collaborative Tasks",
    author = "Bara, Cristian-Paul  and
      CH-Wang, Sky  and
      Chai, Joyce",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-main.85",
    pages = "1112--1125",
    abstract = "An ideal integration of autonomous agents in a human world implies that they are able to collaborate on human terms. In particular, theory of mind plays an important role in maintaining common ground during human collaboration and communication. To enable theory of mind modeling in situated interactions, we introduce a fine-grained dataset of collaborative tasks performed by pairs of human subjects in the 3D virtual blocks world of Minecraft. It provides information that captures partners{'} beliefs of the world and of each other as an interaction unfolds, bringing abundant opportunities to study human collaborative behaviors in situated language communication. As a first step towards our goal of developing embodied AI agents able to infer belief states of collaborative partners in situ, we build and present results on computational models for several theory of mind tasks.",
}

Installation Instructions

This README assumes that the user is about to set up the MindCraft task on a to-be-newly-created Ubuntu-based AWS EC2 Server. If not, some commands may be invalid (e.g. apt-get vs. apt). This has been tested on both Ubuntu versions 18 and 20.

Server Setup & Port Forwarding (for reference, AWS port forwarding guidelines are adapted from here):

  1. Launch an EC2 Instance, choosing an Ubuntu-based x86 Amazon Machine Image such as Ubuntu Server 20.04 LTS (HVM), SSD Volume Type.
  2. Choose an instance type. Minimum resource requirements are relatively high, seeing as we are going to run web & game servers concurrently on the same machine. Testing has indicated that instances like t2.xlarge with at least 16 gigs of RAM work fine.
  3. Leave the options specified in 3. Configure Instance, 4. Add Storage, and 5. Add Tags as default.
  4. On 6. Configure Security Group, Add Rule of type Custom TCP Rule, with options Port Range: 22565 and Source: Anywhere. This ensures that players can access the game server with just the IP address.
  5. Add another rule, this time with option Port Range: 8080, keeping all other options the same as above. This ensures that players can access the web server with just the IP address.
  6. Review and launch, specifying your EC2 KeyPair for remote admin access.

Required Depenencies (install these sshed into the EC2 machine):

  1. Java (for reference, Java-Spigot guidelines are adapted from here)
    1. First, update your local package lists with sudo apt update; this updates the URL locations for all the required dependencies you're going to install later. On a newly-created EC2 server, these lists at startup are going to be horribly out of date.
    2. Next, install Java Runtime Environment 8 with sudo apt install openjdk-8-jre-headless.
  2. MySQL (for reference, MySQL guidelines are adapted from here)
    1. To install MySQL server, run sudo apt install mysql-server.
    2. Run setup script sudo mysql_secure_installation to configure your newly installed server with authentication permissions. Go through the setup instructions, setting up a password for the root-user, and confirming other security settings.
    3. To confirm that your newly-installed MySQL server is running, run systemctl status mysql.service to manually check.
    4. Now, the password just created in step 2.2 doesn't actually enable remote connections as root (which we want for our game server & web server); here, execute:
      1. sudo mysql
      2. In the MYSQL commandline, run ALTER USER 'root'@'localhost' IDENTIFIED WITH mysql_native_password BY 'password'; (replacing password with a password of your choosing).
      3. And finally FLUSH PRIVILEGES; to write changes to disk.
      4. To exit the MySQL commandline, exit.
    5. To create your MySQL database (where all MindCraft user activity is recorded), enter the MySQL commandline this time with mysql -u root -p, and then entering the password you just created.
    6. In the commandline interface, CREATE DATABASE minecraft; to create, and then SHOW DATABASES; to manually confirm the creation of the minecraft database. Before you exit, run SHOW GLOBAL VARIABLES LIKE 'PORT'; to confirm the port that the MySQL server is running on (by default, it should be 3306).
    7. Finally, set up game server plugin authentication details! Before you proceed: if this is your first time setting this up, the files listed below may not have been created yet. To create these files, go ahead and start the server once with bash startServer.sh. You're going to see a ton of errors appear, but hold out for now! When the server is done spinning up, enter stop to stop the server, and then proceed onto these sub-steps.
      1. Open spigot/plugins/situatedDialogue/config.yml, and change:
        1. mysql_password to what you defined in 2.4.2.
        2. mysql_port to what you just confirmed in 2.4.6 (default is 3306).
      2. Open spigot/plugins/LogBlock/config.yml, and change the same lines as above, this time under section mysql. Remember to also change the MySQL user here to root (default initialization has it as something else)!
      3. Open mean/server.js and do the same for MYSQL authentication credentials in the first few lines.
      4. Open spigot/plugins/AdvancedReplay and do the same for mysql.yml.
  3. NPM (for reference, Node.JS guidelines are adapted from here)
    1. Package lists should already be up-to-date, so running sudo apt install nodejs and sudo apt install npm will suffice; this will install the latest versions.
  4. MongoDB (for reference, Mongo guidelines are adapted from here) These steps are necessary mainly because I have used a cookie cutter MEAN stack setup, even though the underlying web server doesn't really use Mongo at all.
    1. Run wget -qO - https://www.mongodb.org/static/pgp/server-4.2.asc | sudo apt-key add -
    2. Run echo "deb [ arch=amd64,arm64 ] https://repo.mongodb.org/apt/ubuntu bionic/mongodb-org/4.2 multiverse" | sudo tee /etc/apt/sources.list.d/mongodb-org-4.2.list
    3. Run sudo apt update and sudo apt install -y mongodb-org to install Mongo.
    4. Run sudo systemctl start mongod to start the service and sudo systemctl status mongod to check service status.
    5. Finally, run sudo systemctl enable mongod to enable Mongo to start on every server reboot.

File Structure

The MindCraft environment is split into three main modules: (1) initialization, by which task variables (e.g. number of games, complexity of games, and more) are set to customized or default values; (2) the game server, a Bukkit/Spigot-made Minecraft multiplayer server that hosts the game itself and records all in-game interactions; and (3) the web server, a MEAN-stack (mainly, Node.JS) web server that records webpage user interactions, where currently recording of player mental states takes place.

All recording of user data -both game server and through the web server- are consolidated into a local MySQL database, the authentication details of which are to be specifiied in initialization files (or left as default).

Execution

Both the game server and web server are designed to be run concurrently in parallel. To achieve this, set up two tmuxes and run the following sections in separate muxes.

Game Server: Run bash startServer.sh. Edit the parameters passed to the plan generator python script if desired before running the server. If it's your first time running the server, you may have to go the newly-generated spigot/eula.txt and change eula=false to eula=true.
Web Server: Navigate to the mean folder and run npm start. Do this after the game server has been successfully spun up.

Replay

For replaying a specific previous game, make sure that the correct plan file of the logged game, indicated in the format logs/logs.XXXXXXX.plan.json (where XXX is the UNIX time stamp of when the game was played), is copied to the following folder and named as plan_generator/plan.json. After this has been done, start the server with bash spigot/start.command instead of the usual bash file.

Dataset

Data

Models

If you are having trouble accessing the data and models please contact Cristian-Paul Bara at cpbara@umich.edu