Vimeo

Analyzed gender biases in Vimeo’s search and recommendation system and formulated a pipeline with metrics that quantified gender biases in search results. Developed proof of concept learning to rank models, and an internal dataset for training/evaluating machine learning based search. Under the mentorship of Silvena Chan.

The Search and Recommendations team at Vimeo is starting their journey to better search by audting for bias. Specifically, we target the following research questions in my internship:

1. Who are Vimeo users? Which ones use search, especially personal search? That’s when a user searches over their own content like their library of Vimeo videos and folders. It’s the search use case our team is most focused on optimizing for.
2. Does our existing search system exhibit biases, and, if so, what’s the impact?
3. What might happen to any existing bias if we move toward search powered more by machine learning?