Machine learning builds models by finding statistical patterns in training examples. Supervised learning uses labeled examples, such as documents coded responsive or not, while unsupervised approaches find structure without labels.
Performance depends heavily on whether the training data resembles the data the model will encounter in production.
Alternative Names:
ML, Statistical Learning
Why it Matters?
Machine learning underlies the e-discovery tools courts have accepted for over a decade, which gives legal buyers a useful precedent: technology-assisted review was validated through statistical sampling rather than through faith in the algorithm. The same evidentiary posture applies to newer applications, and it is why validation on representative documents matters more than vendor benchmarks.
Frequently Confused with
Related terms
Frequently asked questions
How is machine learning different from rules-based software?
Why does training data matter so much?
