Now that you have your data structured and mapped in Pipeless, you can run real-time algorithms on that data. Depending on the data schema you have, there are some algorithms that will perform better for you than others.
Data Schema Considerations
If you have following/friend/connection style relationships in your product then you'll be able to make use of the activity feeds algorithms as well as the recommendation algos. Without any explicit following relationships in your product, you can still utilize most of our recommendation algos as well as the Recent Activity feed endpoint.
Having tags or category data associated with your content opens up more power of Pipeless, but without tags you can still do a lot with the collaborative filtering that powers most of our algorithms. The use of tags in all of our endpoints is optional.
If you don't currently have tagged content in the form of tags, categories, hashtags, genres, topics, etc. you can run some processing to generate tags, like IBM Watson NLP Classifier or AWS Comprehend to turn unstructured text into structured tags. Similar services exist for images, video and audio content that could be used to generate tags for your content.
Activity Feed Algorithms
Get recent user activity related to a specific item (account, product, tag, post, video, etc.). For a target object, this algorithm pulls the most recent user actions on that object.
Get a timeline of content from accounts a user is following. This algorithm navigates the graph of recently posted content from accounts a user is following and combines them into a single feed.
Following Action Feed
Get a feed of activity using multiple signals of user engagement from accounts a user is following. This algorithm aggregates actions from across other accounts a user is following, to bubble up content that those accounts have engaged positively with.
Get a personalized set of recommended content based on a user’s interests and what similar users have engaged with. This algorithm utilizes collaborative filtering based on selected engagement signals and categories/tags for a wide variety of recommendation formats.
Get a personalized list of accounts to follow based on a user’s interests and who similar users follow. This algorithm combines collaborative filtering of similar following behavior with tags from content that a user has engaged with positively in order to provide recommendations of other accounts.
Get similar tags or categories based on shared qualities. This algorithm uses collaborative filtering to find other tags that have been similarly engaged with by other users.
Get content similar to other content based on categories or tags along with user engagement. For a target object, this algorithm uses associated categories/tags along with collaborative filtering from selected user engagement signals to find similar content.
Get a list of accounts who are followed by similar people and have similar interests as the user being viewed. For a target account, this algorithm combines collaborative filtering of similar following behavior with tags from that account's content to find similar accounts.
Learn with the demo dataset
You can also get a feel for our different algorithms by adding our demo app and experimenting with the restaurant demo dataset. Check out the restaurant demo dataset tutorial to walk through all of our algos.
Updated 26 days ago