Siqi Wu, from the ANU Research School of Computer Science, is one of four PhD students in Australia to be selected for the 2018 Google PhD Fellowship program.
The fellowship program aims to support PhD students in computer science and related fields, and encourages innovative research in Australia. Fellowship recipients have the opportunity to build connections between industry and academia, while having access to a Google mentor during the twelve-month program.
We caught up with Siqi to ask him about the fellowship and learn more about his data-driven research that investigates the popularity of online videos.
Congratulations on the fellowship Siqi. Please tell us, how did you become interested in the popularity of online videos?
Personally, I am a video addict and before I was exposed to YouTube, I spent a lot of time on online learning platforms to explore things like programming, statistics, or even some introductory finance courses.
The popularity of online videos and their consumption patterns reveal important things on how human attention is allocated to web context. A billion hours are spent watching YouTube every day, which is 100 years in total. This is why understanding video watching behaviours is of great interest to both the end user and service provider.
The research I am undertaking is very practical and is able to benefit many applications such as recommender systems, system design and others. The Computational Media Lab at ANU has worked on YouTube over the past four years and successfully linked it with other social media sites.
When I started my PhD I was extremely excited with, but felt challenged by the huge amount of heterogeneous data. Now, I feel there are many interesting observations beneath the iceberg that I’m motivated to discover.
What has been the most interesting thing you’ve learnt about online behaviours so far?
I have learnt that online behaviours can be easily manipulated. Obvious examples are clickbait eye-catchy news titles and the more sophisticated social bots. However, there is also a system-level mechanism that guides the attention flux, such as “guess you like" products on Amazon and “recommended for you" videos on YouTube. This suggests for an online item, popular or not, may be determined even before the item is exposed to the public.
What is the surprising thing you have learnt about yourself during your PhD?
As an endurance runner, I have always liked the long and tough activities. I have run around the Fuji Mountains in Japan and have hiked through the wilderness of the John Muir Trail in California but I still find my PhD journey has been the hardest – or maybe I can accredit it for pushing me to the next level.
What do you hope to gain from this Fellowship and what would you like to do when you finish your PhD?
I’m looking forward to connecting with industry, and Google is very appealing! I also look forward to working on new and exciting projects, and maybe an internship sometime next year.
For my future, there are still many possibilities and I may either stay in academia or go back to industry. For now, I will focus on my work and enjoy life in Canberra.
Ok, and the question that everyone wants to ask… Are cat videos and funny fails really the most popular videos on the internet? If not, what is?
They are indeed very popular, who doesn’t love cats and epic fails! However, I would not accredit them as the most popular ones.
There are many metrics to measure popularity and people often overestimate the video view count. Engagement and longevity are also important factors in assessing popularity. While cat videos and funny fails succeed in attracting a lot of attention when first uploaded, people tend to forget them quickly or they are replaced by new uploads.
Accounting for all aspects it is music videos, especially from big artists, that excel in the dimensions of popularity, engagement and longevity. After all, you just need to look at the YouTube music video “Gangnam Style” and the most recent YouTube oligarch “Despacito”.