Challenging issues of automatic summarization : relevance detection and quality-based evaluation