Time: 10 Weeks
Date: Spring 2015
Tools: C++
Source Code: Github
Research Paper: Download
Description:

For my master thesis I researched, prototyped and tested a plugin for Unreal Engine 4 that attempts to provide intelligent suggestions to the users within the blueprint environment. Suggestions systems in code editors have generally been accepted to increase productivity, however, very little work has been done in the field of applying such a suggestion system in a visual programming environment. This research was performed to test whether productivity benefits from providing context sensitive suggestions during the editing process within a visual programming environment.

This research explores the field of suggestion algorithms found in modern implementations, from query auto-completion to graph based completion systems. We then provide the details of the implementation we used to construct the prototype. The prototype was then used to test the two hypothesise defined in the research document: 1) a suggester system will improve productivity in a visual programming environment and 2) context is important to provide accurate suggestions in a visual programming environment.

From this research we have found that using our suggested method we are able to provide an accurate suggestion in a 5-entry short-list in 73% of situations. User tests have shown that perceived productivity increased when using the implemented recommender system, however, timings taken from the same user tests showed that actual productivity slightly decreased.

The source code for the plugin and test project can be viewed here. The plugin requires a modified version of Unreal Engine 4 as it relies on some custom hooks for better integration of the suggestions in the existing systems.