04/10 - Superlinked Framework Update

[ClaireSuperlinked]

We have been busy building the Superlinked and would love your feedback. Share your thoughts and any features you want to see in the Discussions HERE.

🚀New Feature Releases

• Categorical Embedding Space - Allows for embedding categorical data using multi-hot encoding to create connections based on a category instead of semantically. (example here)
• Custom Space - Allows you to parse your own embedded data into the system without requiring a pre-defined space for that data type (for example, images) so that you can experiment with more data types while we build out new Spaces. This embedded data will then be treated as an entity like any other space, and you can apply custom weights etc, to it (example here)
• Performance improvements: If you switch on GPU you can 10x speed for text similarity space ingestion.

🙏 Kudos!
👏 Thanks to maruthiprithivi gauravnbcu murraybutler dhruv-anand-aintech merlin-quix for submitting feedback on VDBs you want to see supported
👏 Thanks to jhk753 for helping us improve performance by spotting a bug in the system

Got a project idea you’d like to try? We’d love to see more of what you’re building with the library. Please share any notebooks you build with the community here

⏰ Upcoming

• README updates to give more details on specific features and a series of article documentation to help explain how the different features work and Getting Started.
• Superlinked-preview: a containerised version of Superlinked that allows you to use the framework through a REST interface with connectors supporting Redis & MongoDB VDBs for initial release.
• Additional use case notebooks to help you explore the functionality with different datasets