NIST had hired human annotators who watched the training movies
and worked to extract and annotate open domain questions and answers
covering various aspects about characters, their actions, and events
during the movies. Questions varies in complexity and objectives (e.g.
low-level computer vision goals, to high-level semantics).

This folder includes a text file per training movie. Each line consists
of 1 question followed by multiple answer options and ends with the correct answer
in the following format:
<Question text>:<option 1>:<option 2>:<option 3>: .... :<correct answer>