VID2XML: Automatic Extraction of a Complete XML Data From Mobile Programming Screencasts

Developers often refer to video-hosting online platforms to find screencasts that provide a step-by-step guide to help them solve a programming task at hand or learn a new concept. More specifically, developers search for resources that help them design and implement effective mobile graphical user interfaces (GUI) using
XML
. Although mobile programming screencasts contain a vast amount of
XML
data at developers’ disposal, they cannot be easily found and copied-pasted due to the image nature of videos. Given that the most common task developers perform online is copy-pasting, mobile programming screencasts must support that and be complemented with
XML
data in a textual format. To overcome this challenge and aid developers, this paper presents
vid2XML
, which is a three-phase approach that leverages both visual and textual information of video frames to locate
XML
region in video frames, locate the currently opened file, and extract
XML
data for each file presented in video frames. We evaluated each phase of
vid2XML
in a comprehensive empirical evaluation on videos collected from YouTube. The results reveal that
vid2XML
is able to accurately (i) locate
XML
regions, outperforming four previous work, (ii) locate the bounding box of the selected file, and (iii) extract, fix, and merge
XML
data for each file opened/created in a video.