網際網路的發展，資訊量快速成長，資訊過載問題日益嚴重，為了能有效率管理 龐大的資訊，資料須適當的處理，幫助使用者整理龐大的資訊並加速獲得真正有用的 資訊。傳統的文件分群主要使用字詞在文件中的權重當向量空間模型的依據，得面臨 一些挑戰，如：資料量大時，高維度向量稀疏矩陣需要大量計算成本且效能不佳、詞 彙為獨立構成，無法區分文中詞彙間關聯性、並不是所有詞彙一樣重要。本研究提出 一套方法，透過分析字詞與字詞間連結度，形成字詞群集，利用字詞群集協助文件分 群。首先，針對資料集擷取資訊量較多之關鍵字當字詞群集之基礎；接著，依關鍵字 平均連結度分數加以合併形成字詞群集，用以表達文件進行分群。由實驗結果顯示本 研究提出之方法能提升分群之效能，更能夠表達詞彙在資料集與詞彙之關係。 The World Wide Web continues to grow at an amazing speed to bring a quickly growing number of documents. Since information overload is more serious than ever, the development of new methods for managing these information is an important issue. In most document clustering algorithms, documents usually are represented in the vector space model, which consider all dimensions (terms) in similarity measurement. In this vector space model, there are some weaknesses. First, cost much in calculation in high dimension situation. Second, it treats terms as independent and of equal importance. In this paper, we propose a method to aid document clustering. To start with, we analyze degree of word connectivity and then, group keywords in to keyword clusters finally, all documents were clustered according to the score among the keyword clusters and then choose the highest score keyword cluster for each document. Our experimental results show that the performance of the proposed approach has been improved effectively.
