事件熱門趨勢分析與部落格文章推薦機制之整合研究

Abstract

近來部落格(Blog)儼然成為Internet 上新興的溝通以及發布內容的媒介，許多最新與有 用的資訊皆透過其擴散，網路上因而充斥著大量的部落格文章，使得部落格讀者面臨資訊 過載的問題，因此，部落格文章推薦之研究議題愈趨重要。部落格文章通常包含有事件相 關的資訊，透過分析落部落格文章可分析出事件的相關趨勢，而使用者通常會透過部落格 來閱讀有興趣的熱門事件文章，然而，現有部落格文章推薦之相關研究並未探討個人化的 熱門事件文章推薦。傳統部落格事件分析之相關研究著重於分析包含特定關鍵字詞彙文章 數量以偵測事件之發生，並未預測其熱門程度之發展趨勢。此外，熱門事件通常會吸引大 量的使用者透過Google 搜尋引擎搜尋與事件相關的資訊，而目前尚未有相關研究整合 Google 關鍵字搜尋資料予以分析事件的熱門程度。 本計畫擬結合部落格文章與 Google Insights 為基礎之事件分析，發展創新之預測事 件熱門趨勢方法；此外，部落格讀者對於不同的熱門事件有其個人之偏好，本計畫將探討 如何設計推導個人化熱門事件偏好之方法；以此為基礎，本計畫將設計研發整合熱門事件、 文章熱門程度與推薦機制之個人化部落格文章推薦方法；此外，群體偏好分析可改善個人 偏好分析之不足，本計畫將進一步結合群體偏好分析，設計研發以個人化調適性權重調整 群體與個人偏好相對重要性之部落格文章推薦方法。 本計畫提出具學術創新的研究構想，將研發新穎的事件熱門趨勢預測方法，以及新穎 的部落格文章推薦方法，並擬以使用者實際的部落格文章瀏覽紀錄進行方法的實作與實驗 評估，本研究預期對於學術創新以及部落格文章推薦之實務應用有所貢獻。

Recently, weblogs have emerged as a new communication and publication medium on the Internet for diffusing the latest and useful information. With abundant blog articles published on the Internet, blog readers are likely to encounter with the problem of information overload. Accordingly, blog article recommendation has become an essential and emerging research topic. Blog articles usually contain event-sensitive information, and thus the event trends can be obtained through analyzing blog articles. Blog readers are often interested in articles that are related to popular events. However, existing researches did not address the issue of recommending blog articles of emerging or popular events that suit user interests. Previous studies on analyzing blog events focused on identifying hot events by analyzing the popularities (occurrence frequencies) of event terms in blogosphere, and did not predict the popularity trends of events. In addition, a popular event usually allures a large number of users to search information related to the event through Google search engine. However, no research has been devoted to integrate the analyses of blog data and the search data of Google Insights to predict the popularity trend of an event. In this project, we will develop a novel event-trend analysis approach to identify events and predict the popularity trend of events by integrating the blog-based and Google-based popularity trend analyses. Moreover, blog readers usually have different interests in emerging or popular blog events, and thus we will investigate how to infer their personal preferences on popular events. This project will develop novel personalized recommendation methods for recommending blog articles of popular events by integrating recommendation methods, the popularity trends of events and article popularities. In addition, group preferences usually can complement the deficiency of personal preferences. Thus, this project will further investigate the methods for analyzing group preferences on popular events, and develop an adaptable weighting mechanism to adjust the relative importance of group and personal preferences for blog article recommendations. This project will propose novel research ideas and develop novel approaches for predicting the popularity trends of events and recommending blog articles. The implementation of proposed approaches and experimental evaluations will be conducted by using real data obtained from blog websites. This project is expected to have contributions to academic innovations and practical applications of blog article recommendations.