隨著網際網路及資訊技術普及應用，人們經由網路所獲得資訊的時間越來越短，但 資訊量急劇增加，過量的資訊逐漸形成資訊爆炸的問題，各個企業與機構的數位文件也 不斷快速累積，數量大到難以有效的管理與利用，文件分類 (Text Classification) 因應而 生，利用自動化的技術協助人工分類，來應付大量暴增的分類需求。傳統文件分類的做 法是以人工方式進行，近年來，深度學習 (Deep Learning) 已被廣泛討論並應用在多種 研究上，許多文獻顯示利用深度學習的技術，可以幫助結果更加完善或增進效能。 本研究利用消費者於實體店面購買的商品名稱資料，透過新興的深度學習技術，應 用 word2vec 詞向量模型於文件自動化分類，藉由其自行學習語義間關係的技術，將商 品自動分類到正確的類別，並透過四個實驗探討不同的因素下訓練出的 word2vec 詞向 量模型，會影響其成效，最後也證實以 word2vec 擴展關鍵字詞能提高分類成效。;With the popularization of Internet and technology, people get more information through the Internet, but the amount of information has increased dramatically. Excessive information has gradually formed the problem of information explosion. Digital documents of various enterprises and organizations are also constantly increasing. The amount of digital documents is large to be difficult to manage and utilize effectively. Text Classification is created in response to deal with the massive surge in classification needs. Traditional text classification is done manually. In recent years, Deep Learning has been widely discussed and applied in a variety of studies. Many literatures show that deep learning techniques can help improve results or improve performance. This study uses the data of product names purchased by consumers in physical store to apply the word2vec word embedding model to the automatic classification of documents through the deep learning technology. By self-learning the semantic relationship, the products are automatically classified into the correct category. And through a number of experiments to explore the word2vec word embedding model trained under different factors, will affect its effectiveness. Finally, this study confirmed that applied word2vec to expand keywords can improve the effect of classification.