recommendation engine

From IndieWeb
Jump to: navigation, search


A recommendation engine is software or a service that recommends things to people based on their behavior or people that have behaved like them in the past, like the features of sites that say things like people that bought this also bought that, or liked, or followed, etc. Such recommendations are also known as targeted resource recommendations, and are usually in a specific domain.

Why

Reasons to have a recommendation engine in your site / service / software include:

  • To improve search (rather than relying solely on keywords)
  • To suggest content after onboarding (rather than presenting the user with a blank or template page)
  • To help users follow each other (and therefore stimulate more conversation)

IndieWeb Examples

None currently.

Silo Examples

Amazon

Amazon recommends things to buy based on:

  • What you've bought in the past
  • What others have bought that you have bought
  • What others have bought after looking at the same page as you're on
  • ...

Facebook

Facebook has a few features that appear to use a recommendation engine:

Netflix

Netflix recommends movies to watch based on:

  • What you've watched and others have also watched
  • ...

Twitter

Twitter has several features that appear to use a recommendation engine:

  • who to follow
  • suggested accounts when creating a new account (screenshot?)
  • suggested accounts to follow shown immediately after following someone (screenshot?)

Services to Silos

Tumblr

Tumblr recommends other blogs you should follow presumably based on blogs you already follow.

Also in the sidebar is a section called "radar" which shows and links to a single post.

tumblr-recommendations.png

When you click on Explore (the compass icon) you also get a list of topics and posts Screen Shot 2018-12-04 at 12.53.11 PM.png

When you click on a hashtag on Tumblr you get a list of posts plus related hashtags Screen Shot 2018-12-04 at 12.54.32 PM.png

In any of the recommendation tools you can sort by recency or by post types. Screen Shot 2018-12-04 at 12.55.38 PM.png

Screen Shot 2018-12-04 at 12.55.32 PM.png

Brainstorming

A recommendation engine might recommend books based on titles you've already read, or blogs based on who you're following.

In an IndieWeb context, a recommendation engine could recommend

  • other independent website owners to follow an indie reader, based on
  • articles to read, based on
    • articles liked by people who have liked other articles that you've liked
  • liked/bookmarked/etc posts based on similarity towards the current entry; maybe with tag matching, text analysis. deep learning or manual sorting
  • YaCy might be used for indexing and for knitting together distinct communities that are scattered across many social networks
  • ...

One of the benefits of an indieweb approach is that you might be able to supply your own algorithm, or have a choice of algorithm, rather than needing to trusting one provided by a silo.

The benefit might be a wider audience for independent site owners whose content might not otherwise be discovered.

With Microsub it would be possible to build a microsub "middleware" app that could keep track of what content is popular among its users and automatically create feeds of recommended content without needing to provide a full ui

See Also