The architecture of the tool contains two main components. The components were structured based on two essential purposes:
We have provided our visualization tool to four developers who participated in our study. The developers had the independence of not choosing the visualization tool for careful inspection of the resultant dataset, but all of them found the visualization tool to be highly conducive for the exploration of the result set.
We have also invited 4 of the undergrad students (who had the basic knowledge of machine learning and were paid for the tool-evaluation) to find out the perspective of fresh/new users. All of them reported that visualization tool to be robust and engaging.
One of the student participants pointed out the lack of options presented in the topic modeling part, which we found to be true. Nevertheless, at the same time, the comprehensiveness, smooth interactability, aesthetic features, and user-friendliness of the tool was also highly commended.
Before implementation, we have researched different ways to find a convenient and concise way of presenting the generated data so that developers could compare and utilize the extracted information with maximum efficiency. We have successfully delivered an interactive visualization tool that would be constructive and supportive for the developers who frequently analyze and inspect user feedbacks to identify energy efficiency issues.