本研究では,genericでinformativeな要約の作成を目指し,「指定された要約率の範囲で,元のテキストの情報をできる限り含めること」と「作成された要約が文章として自然で読みやすいこと」の実現のための重要個所抽出による要約作成を行なう.本稿では,重要個所抽出の際に,構文情報と語彙的結束性の情報を考慮して,同じ内容を表わす語の繰り返しによる冗長性を抑える手法,文としての意味を維持するために必要な他の部分を補完する手法,内容的に一貫性のある読みやすい要約を作成する手法について述べる.また,指定された要約率の範囲で内容がどの程度保持できているかを人間による重要個所抽出型要約との比較で評価する.In this research, we aim to develop a summarization system which is capable of making an informative and generic summary. We require the system to have two abilities; to include as much information as possible on the original text in the summary of the specified rate, and to produce natural sentences with higher readability which can be exchanged for the original text. We adopted the method of extracting important parts of sentences for our first summarization system. Since important parts are smaller than important sentences, it can be considered that a fine-grained summary can be produced by using these parts. In this paper, we describe an automatic summarization method of extracting important parts which result in an improvement of readability and an exclusion of redundancy. We also make a content-based summary comparison by measuring the similarities between the summaries extracted by some automatic summarization methods and those extracted by human subjects.