What is a primary benefit of creating a dedicated "knowledge base" for a legal chatbot, as opposed to solely relying on web scraping for content?