When your only tool is a hammer, everything looks like a nail. One of the most frequent questions I get asked about Natural Language Processing (NLP) is which is best ‘rules’ or ‘machine learning’. With the popularity, hype even, of Large Language Models (LLM) it would seem its obvious – machine learning because of its ability to generalise?
The analogy I’ve been exploring is the car (machine learning) and the bike (rules).
1. Few would argue a bike can out perform a car over almost any distance.
2. Few would argue a car is more nimble or manoeuvrable as a bike.
3. Few would argue against the fact a bike is cheaper than the average car.
4. Few would argue against the fact a car can carry more weight than a bike.
5. Few would argue against the fact a car has greater startup costs, needs fuel before it can start, is more complex to maintain and needs more training to use safely.
6. Few would argue against the fact a car can take us to many places that a bike cannot, although a bike can get to some a car cannot.
Sometimes the bike is best, sometimes the car, and sometimes you can put the bike in/on the car and use in conjunction.
They co-exist. Understand the business opportunity or problem, then pick the best tool(s) for the job.
I came across the article below, that is well balanced, covering pro’s and con’s for each. Worth a read…
A Geodyssey - Geoscience Text Analytics and Enterprise Search Research 
Leave a comment