In developing a solution utilizing the Language service, which natural language processing technique is best suited for extracting the primary topics from a set of documents?