Working on the OpenCog codebase requires a good understanding of the theory. The OpenCog By Way of Wikipedia page provides some of that. Reading Ben Goertzel's book, Real World Reasoning won't help much with the current codebase, but it will give you an overview of the the OpenCog project.
The correct way to get started depends on your interests.
- Scientists interested in the relationship between neural nets, deep learning, distributional repesentations and symbolic AI should take a look at the unsupervised language learning project. It attempts to explore and demonstrate how these different approaches are actually opposite sides of the same coin. Here's a a mathematical explantation and here's a non-matthematical overview.
- Computer Science professionals interested in knowledge representation and programming languages should take a look at Atomese. This is a programming language designed not for humans, but for machines, providing a sophisticated, powerful API to knowledge representation and manipulation tasks, including learning and reasoning. It's a mashup of ideas from data mining to functional programming, from robot control to SAT solving, from ProLog to relational query languages. Its built on the AtomSpace, which is a graph database with advanced features not found in other graph databases.
- If you are a student, take a look at the ideas page. There are many different kinds of projects described there: things worth understanding and working on. Students are encouraged to "drill down": pick a specific task, and dive deep into it, reading and studying everything you need as background.
- Software developers are encouraged to take the opposite tack: get a broad overview of the system, the software, the components, what they do, and how they work. One of the best ways to get started is by fixing bugs. There's a bunch to choose from, on github.
- If you are a systems integrator, wishing to build sophisticated, complex systems, take a look at ghost - this is an avatar control system, designed for integrating robot control with natural language, common-sense knowledge and reasoning. Oh, and scripted chat. Mostly that, as scripted chat remains central to usable avatar systems.
- For supervised machine learning, MOSES remains a solid and useful system.
If you need additional assistance, feel free to ask for help.
Hands On With OpenCog
The Hands On With OpenCog tutorial will walk you through the steps to build and start using the AtomSpace. An obvious first step is to get a copy of the software. Instructions for obtaining source code, getting the correct dependencies, and building the software are available at Building OpenCog. Don't forget to study the theory of the AtomSpace and Atomese.