mismo 0.1.0

No project description provided

Mismo

---

Table of Contents

• Goals
• License

Goals

Use Ibis as the core

This gives a few benefits that are key to record linkage:

• Ability to use datasets that are larger than memory
• Ability to use multiple backends (eg duckdb for single node, or bigquery or spark for distributed)

Thoughtful, composable API

Use a duck-typing approach to allow users to plug in their own components eg "Blocker" has a block method with a certain signature. This makes mismo a bit more complicated than dedupe or splink, but it will be much more flexible.

Extras

• More ergonomic model persistence than dedupe. splink did a good job here.
• Support determinism using random_state (unlike dedupe)

License

mismo is distributed under the terms of the LGPL-3.0-or-later license.