Annif, ranking, and trying out hybrids
· 7 min read
Subject headings - using ANNIF, and the various techniques we used to make it more useful (some more successful than others).
lentil party
Storm-powered* local AI
· 3 min read
Fourteen thousand books catalogued on wind power, for the carbon cost of boiling a cup of water.
Adding confidence scores with system one models
· 7 min read
The model can't tell us when it's unsure, so we've been building checks from outside it: some plain code, and some models that only answer yes or no.
Through a glass darkly
· 10 min read
We asked a local model to explain its decision. The explanations had nothing to do with the decisions.
Introducing Paratext
· 7 min read
Making an images-to-output workflow that is simple, extensible, portable, and intuitive.
Reaching good enough
· 5 min read
Reaching a 'good enough' threshold on model-generated metadata
Fine tuning a model on the local machine - powered by wind
· 2 min read
Fine-tuning a model entirely on the local machine, and making a tool to do this as sustainably as possible.
Why the Framework Desktop
· 5 min read
Why we chose the Framework Desktop as our local AI platform.
Beginning the work
· 3 min read
How we think about AI, and the beginning of a local journey.