📜 ⬆️ ⬇️

Expansion of the base for collaborative filtering

The idea of ​​collaborative filtering is simple and elegant - based on the list of user preferences, the system searches for people with similar preferences, compares lists and issues recommendations for their completion. The word “preference” is not quite appropriate here, usually the list contains the names of objects of some content type - books, music, films. But if we allow a blog platform or social network is looking for your friends? Then, the user’s interests specified in the profile, tags to his posts, a list of existing friends, etc. can serve as elements of the list. If you aim to maximize the scope of collaborative filtering, two questions arise: what can be a list element and which, in principle, can encourage the user to create these lists.

Regarding the first, in my opinion, the maximum abstractions were achieved in the Metaweb project with their Freebase database, recently purchased by Google. This database is a graph whose nodes are arbitrary entities or entities — people, places, things, organizations ... For example, at the request of “site owner” it is issued: “this is an entity that owns and / or manipulates the site” ( A website owner is an entity that owns and / or operates a website ). The websites themselves are also entities. They are the interests, themes, concepts. A user who would list all these objects in his personal "megaspill" would give a wealth of information for analysis in the systems of collaborative filtering.

To the question (second) why he should do this, the recommendation services like imhonet are answered simply - the user understands that after creating the list he will be advised of something useful. But in this case it will not work - the list will be too huge (for that matter, my first and last experience of acquaintance with the literary department of imhonet alone was tedious - I just forgot to enumerate the books).
A possible solution is the “Like” button, which you can click when you hover over a selected object, for example, when you select a word from the text. Or in a more advanced form with some kind of visual recognition of objects in the real world using smartphones. I saw a girl, clicked Like. This activity is not particularly stressful and not one-time, you accumulate your list of like-objects gradually, in the process of surfing the Internet and familiar places. Although I personally prefer the verb "friend". If we are frend people, why not frend objects?

Another possible option is to combine communication with the creation of graph databases of the type mentioned. In this case, posts and comments will also become objects of a common network and users will generate connections between objects in the process of familiar activity such as communication and networking. Those. This has nothing to do with collaborative filtering, but some of these links will be useful in terms of its application. This type of service I advocate .

Source: https://habr.com/ru/post/101631/

All Articles