YouTube recommendations have always been a point of discussion. Some users find them useful, others are not convinced about their relevance. Yet, they exist and need to be talked about. As a lot of these recommendations decide what YouTube users eventually end up consuming. YouTube believes that it is really simple to diversify recommendations, but adding value to them can be challenging. In simple words, YouTube can show you content from 100 different channels but, you may not want find value in all of them or will want to consume all these videos.
"We definitely want to diversify recommendations on YouTube. It actually benefits all parties involved, but it’s not easy. We can simply start recommending videos from different channels and accounts, but, users don’t want that sort of diversification. They might not find these videos useful. They want recommendations that they are actually interested in," Cristos Goodrow, Vice President of Engineering at YouTube told India Today Tech, during a media briefing.
YouTube, on its recommendations page, explains that it takes a lot of factors into consideration, including the watch and search history. It also considers context, such as your country and time of day, which helps the platform show you locally relevant news".
There are times when users click on a video by mistake or watch a few seconds of one before realising it's not relevant. The way the algorithm works, they get recommendations from the same channel. Cristos said that this is one of the challenges YouTube has faced but, it has also learnt through the process.
"We have been trying and have learnt through the years. More than one metric is now considered to ensure that users get diversified and personalised recommendations," he said.
Another factor that YouTube's recommendation systems consider is whether others who clicked on the same video watched it to completion or not. YouTube believes that this is a sign that the video is higher quality or enjoyable. It also looks to improve recommendation systems using random surveys that appear on their homepage and elsewhere throughout the app.
But, what if a user hasn't signed into their account or is completely new to the platform? First of all, users need to understand that YouTube treats new users separately from those who are yet to sign in. If someone is using the account without signing in, they will still be getting recommendations based on what they have consumed.
"YouTube recommendations are usually based on what people are watching. We consider click, watch time and several other factors for providing recommendations. However, when we don’t have much information about the user, it is then we provide general recommendations," Cristos explained.
No comments:
Post a Comment
Welcome To My Blog.