Maps and Models
1 2023-01-04T08:02:48+00:00 Jotsna Iyer 4f2bfb514a09301de0e5275ee45bf5db41479839 7 1 plain 2023-01-04T08:02:48+00:00 Jotsna Iyer 4f2bfb514a09301de0e5275ee45bf5db41479839This page is referenced by:
-
1
2023-01-04T08:02:47+00:00
AI Speak : How Youtube Learns You Part 1
3
plain
2023-02-08T19:03:15+00:00
Models and Recommendation
ACTIVITY
These are the credit card transactions of two people living in Nantes. They are looking for things to try this weekend. What will you recommend to John Doe and what to Tom Harris?
List to choose from :- The new Burger King outlet
- An Olive oil tasting event
- An Online Luggage store
- A river-side Concert
- Baby swimming class
Recommendation systems have been around atleast as long as tourist guides and top ten lists. While The Guardian Best Books of 2022 recommends the same list to everyone, you would likely adapt it when choosing for youself : pick a few and change the order of reading based on your personal preferences.
How to recommend options for strangers? In the activity above, you probably tried to imagine their personalities based on the given information: you made judgements and applied stereotypes. Then, once you had an idea of their type, you chose from the list things that could (or not) be relevant to them. Recommenders like Amazon, Netflix and Youtube follow a similar process.
Nowadays, whenever someone is searching for information or looking to discover online content, they use some kind of personalized recommender system.1,2 The main function of Youtube is to tell its users what to watch amongst all the videos available on the platform. For signed-in users, it uses their past activity to create a "model", or a personality type. Once it has a model for John, it can see who else has models similar to him. It then recommends to John both videos similar to what he has watched and those similar to what others like him watched.What is a model?
Models can be used to mimic anything from users to videos to lessons a child has to learn. A model is a simplified representation of the world so a machine can pretend to understand it :How Youtube learns you
All recommendation problems involve a surrogate question: "What to recommend" is a bit too general and vague for an algorithm. Netflix asked developers what will be the rating a user A would give video B, given their ratings for other videos. Youtube asks what the watch time would be for a given user in a particular context. The choice of what to ask - what to predict, has a big impact on what recommendation is shown.3 The idea being that the correct prediction will lead to a good recommendation. The prediction itself is based on other users with a history of similar tastes.4 That is, users whose models are similar.User models
Youtube splits the task of recommendation into two parts and uses different models for each.3 We, however, will stick to a simpler explanation here.
For creating a user model, its developers have to ask themselves what data is relevant to video recommendation. What about what the user has watched before? What has been their reviews, ratings, and explicit preferences thus far? What did they search for? And more than these explicit signals, Youtube uses more of the implicit kind, since they are more readily available3 - Did a user just click a video or did they actually watch it? If yes, for how long? How did the user react to previous recommendations?1 Which ones he ignored? Apart from the answers to these questions, demographic information like gender, language, region and device are of great value when the user is new or not signed in.3
Once a model is available for each user, we could see which users are similar to one another and use that information for recommendation.
Video Models
Similar to users, we could also use the videos that are similar (or different) to each other. Given a video, Youtube looks at the content it has, its title and description, video quality, how many people have watched it (view count), liked it, favorited it, commented on it or shared it, the time since it was uploaded and the number of users subscribed to the parent channel.1
What a user watches next will also depend on whether one video is part of an episode of a series or an item in a playlist. If a user is discovering an artist, he might move from the most popular songs to smaller niches. Also, a user will not click on a video whose thumbnail image is not good.1,3 All this information goes into the model too.
One of the building blocks of the recommendation system is to go from one video to a list of related videos. In this context, we define related videos are those that a user is likely to watch next.3 The goal is to squeeze the most value out of data to make better recommendations.4
------------------------------------------------------------------------------------------------------
1 Davidson, J., Liebald, B., Liu, J., Nandy, P., Vleet, T., The Youtube Video Recommendation System, Proceedings of the 4th ACM Conference on Recommender Systems, Barcelona, 2010
2 Spinelli, L., and Crovella, M., How YouTube Leads Privacy-Seeking Users Away from Reliable Information, In Adjunct Publication of the 28th ACM Conference on User Modeling, Adaptation and Personalization (UMAP '20 Adjunct), Association for Computing Machinery, New York, 244–251, 2020
3 Covington, P., Adams, J., Sargin, E., Deep neural networks for Youtube Recommendations, Proceedings of the 10th ACM Conference on Recommender Systems, ACM, New York, 2016
4 Konstan, J., Terveen, L., Human-centered recommender systems: Origins, advances, challenges, and opportunities, AI Magazine, 42(3), 31-42, 2021
