[Defense] Identifying Important Predictor Variables in Keyword Generation for Lecture Videos
Friday, May 7, 2021
1:00 pm - 2:00 pm
In
Partial
Fulfillment
of
the
Requirements
for
the
Degree
of
Master
of
Science
Farah
Naz
Chowdhury
will
defend
her
thesis
Identifying
Important
Predictor
Variables
in
Keyword
Generation
for
Lecture
Videos
Abstract
Lecture video is an important learning resource. Although video summarization is a long-studied problem, in recent years video lecture summarization has emerged as a new research area. Navigating through lecture videos to find the content of interest is time-critical for this format. Automatic discovery of keywords for lecture video segments can improve the navigation. The suitability of an N-gram to be a keyword depends on a variety of factors including frequency in a segment and relative frequency in reference to the full video, font size, time on screen, and the existence in domain and language dictionaries.
The goal of this thesis is to gain a finer understanding of how various factors contribute to predicting keywords through experimental results rather than trial and error methods so that better systems can be designed. As a preliminary step, we conducted correlation analysis to see whether there is an association between each factor with the ground truth. To measure the cause and effect of various factors through different metrics we conducted binary logistic regression analysis. The analysis for this thesis employs a real-world dataset from Videopoints, a video management portal. The findings from these investigations can help guide future research in this area.
Friday,
May
7,
2021
1:00PM
-
2:00PM
CT
Online
via
MS
Teams
Drs. Jaspal Subhlok and Thamar Solorio, thesis advisor(s)
Faculty, students and the general public are invited.
