Using Rant in my Python program because I’m a glutton for punishment

Over the past few months, I have been researching and developing a little procedurally generated game which will eventually be created in the Godot engine. This game will have a story that’s procedurally generated for the players. A part of this game is the dialogue, which will also be procedurally generated. To accomplish this, I set out to find a library of some kind which can create procedurally generated dialogue (or at least the dialogue that I want) and is written in my programming language of choice, Python. From the looks of it, there isn’t one, and so I had to look elsewhere. That’s when I stumbled upon something called Rant. This is billed as a library which can procedurally generate dialogue. At first I thought I had found what I was searching for. Sadly, though, it is written in  the least open source-friendly language I have ever seen: C#. This can be used on Linux (with the Mono runtime). But I’m looking for a solution where I don’t have to use a bunch of programming languages to achieve what I want.

At first, I tried making some kind of dialogue scheme that would suit my needs. I threw in some sentences of what may define the NPC, and mashed it all together. From the looks of it, though, the scheme is getting out of hand. I have several lines of dialogue, and I’m not even finished. I don’t entirely know how I’ll fit it all together, considering this is just for a simple demo of the full game. It looks like I’m going to have to get creative.

I went digging and searching around I came upon several possible ways of integrating C# code into Python code. There’s IronPython, a fully implemented version of Python in C#. The big problem with this was that it didn’t look very portable to me, as I would have to bundle the .NET libraries with the game for each platform, and that’s a royal pain the ass. Then I looked at Python.NET, which looked very promising: you can call some C# code from Python, and you can call some Python code from C#. It looked like the best of both worlds. Now, actually making it work is a bigger problem.

When I tried to use the Rant.dll assembly in my Python program, I found that I can’t do that because, well, it’s C# code, and the regular old CPython (which comes with many Linux distributions) can only import C or C++ code. Then I looked into using the clr module from Python.NET, but I couldn’t find a version built for Linux. Through a lot of hand wringing, brow beating, and code cracking, I found that I had to use the latest version of Mono (version 5.0.1) along with an unstable version of Python.NET. This one built with the suggested command: python build_ext --inplace. The built shared object library file, “”, and the “clr” module load in Python.  Heck, I was even able to load the pre-built “Rant.dll”. But this is nothing compared what I must do now: actually making some procedurally generated dialogue with Rant. And I don’t know where to begin with that.

Trying My Hand at Making a System for Procedurally Generating Stories

Last year, I got the idea of making a game where a group of friends could get together, have a game scenario generated for them, and they could start playing.  This is something called procedural generation, wherein game settings, mechanics, assets, and other components are created from an alogorthim, and a bit of randomness.  I thought I would just research this, since creating any sort of video game takes a lot of time and skill.  In this case, I was looking into a system for creating stories.

I’ve read a few things about procedurally generating stories (for instance, Tail-Spin and Minstrel) and I stumbled upon one that may be in my reach.  From a paper titled “Random Word Retrieval for Automatic Story Generation” I found out about something called ConceptNet.  This is a commonsense knowledge database for finding the relations between concepts via semantics.  So you can find one concept (like “dog”) and find corresponding concepts for it (such as “animal”).  The paper talks about the intent of making a system for using ConceptNet to find the relations between words (in a process called “Concept Correlation”) and make a story out of it.  Sadly, they haven’t yet implemented the system.  So what I’m trying to do is make something that will, uh, sort of make a story.

In the paper, they talk about how the system requires a Knowledge-Based System.  Unfortunately, this system is quite tedious and difficult to create (or so the paper says).  So I’m just trying to find the connections between words and concepts.  All that I’ve been able to do, though, is mess with the Python interface to the ConceptNet website, and maybe find some related terms.  Finding the connection between two dis-similar terms is difficult, because the concepts have many branching nodes which connect to other nodes.  Finding the right nodes which connect the two concepts would take a while, because the system would have to iterate over the nodes until they find a match.

The examples I’ve been using have been “dog” and “snow”.  So the system would have to go through each node in the “dog” concept until it found a node which connects to “snow”.  It could be any connection from “dog” -> hasA-> “nose”  -> RelatedTo-> “wet” -> PropertyOf -> “snow”.  Please note that these aren’t actual connections in ConceptNet, but something like this can be found in the database.

So I don’t know how I’m going to tackle this monster, let alone make a story out of it.