Live Coding Video
Although the data collection process for the live coding videos was extremely comprehensive, there were parts of the process that could have been more streamlined. One such part was the need to sift through each screenshot so the OCR algorithm was operated on the correct images and the correct parts of each image. Similarly, the manual rebuilding process was often a bit tedious as it required careful inspection of each screenshot. The next steps to improve these procedures would be to write scripts to automate both processes.
Heat Map
With a longer timeline, there are avenues for improvement and future refinement of heat map development. One area of improvement is within the logic section of the creation of the DataFrame, which stores the line number of the code file along with its corresponding number of changes. As seen by the outlier in line 12 of the learner programmer’s heat map, adjustments should be made to the way lines are compared and tracked. This can be done by using a finer threshold for the SequenceMatcher ratio, which would ensure that a corresponding line is not being assigned incorrectly. The creation of the heat map visualization can also be refined. The script currently uses the Python seaborn package to create the heat map from the DataFrame, but a future scope is to visualize the heat map using a JavaScript library called d3.js. d3.js has much greater functionality, flexibility, and specificity than the seaborn package. Additionally, d3.js displays the heat map in an HTML space, which facilitates simple distribution and allows others to view the heat map easily.