How Lucene reduces CPU usage over full-text search of sql server?

I had read article about how stackoverflow reduced its CPU usage by using lucene.

But my question is how ?

Is it due to caching of lucene ?

and if yes. If we implement a sqlserver fts with mem开发者_开发知识库cache. will it be same as lucene ?

or does Lucene uses different data structures for search ?

Lucene is using indexing and full text search - it's more than caching.

SQL is a set-based relational language. It's not built for ad hoc queries of documents. The technology is completely different.