Knowledge shortage is a giant drawback for a lot of knowledge scientists.
Which may sound ridiculous (“isn’t this the age of Huge Knowledge?”), however in lots of domains there merely isn’t sufficient labelled coaching knowledge to coach performant fashions utilizing conventional ML approaches.
In classification duties, the lazy method to this drawback is to “throw AI at it”: take an off-the-shelf pre-trained LLM, add a intelligent immediate, and Bob’s your uncle.
However LLMs aren’t all the time the perfect software for the job. At scale, LLM pipelines might be sluggish, costly, and unreliable.
An alternate possibility is to make use of a fine-tuning/coaching approach that’s designed for few-shot situations (the place there’s little coaching knowledge).
On this article, I’ll introduce you to a favorite strategy of mine: SetFit, a fine-tuning framework that may assist you construct extremely performant NLP classifiers with as few as 8 labelled samples per class.