摘要
我們身在一個資訊容易取得的時代，而面對巨量的資訊時，應如何快速又準確地獲取需要或有興趣的資訊是我們有興趣探討的議題。因此，本論文設計和實作了一個新聞推薦系統，能夠依照使用者的喜好進行新聞的推薦。另外，本論文也實作了當日熱門詞彙推薦，可以透過不同類別的新聞來推薦熱門詞彙給使用者。 新聞推薦系統藉由蒐集使用者的瀏覽行為來掌握其個人的喜好與興趣，進一步運用使用者的瀏覽行為，結合關鍵詞擷取的技術與TF-IDF演算法計算權重，最後依照使用者的喜好進行新聞的推薦。由於我們採用新聞作為推薦的基礎，而新聞往往包含許多流行語或新名詞等，為了提升關鍵詞擷取的準確度，我們將利用引號區塊擷取法與N-gram產生新詞。
並列摘要
Nowadays, a vast amount of information is freely available on the internet. How to acquire needed and interesting information rapidly in the network is an important issue. In this thesis we design and implement a news recommendation system that can recommend users interesting news according to their preference derived from their behavior. In addition, it also recommend hot key terms in each category of news source. In order to get the preference of users, our system collects the browsing behavior of users, extracts key terms in each news article that has been read, and perform statistical analysis to derive the interested set of key terms as user preference. The keyword extraction is based on the TF-IDF technique. As the news articles sometimes may contain new terms not existing in the dictionary, we need to find a way to generate new terms. We use punctuation block extraction method and N-gram to generate new terms.