異質網站推薦優質書評

Abstract

在充斥著相似類型書籍的市場上，讀者們通常是在有限時間之下去挑選書籍來閱讀。在決定要閱讀哪些書籍時，讀者們傾向去參考專家們所寫的書評。然而，優質的書評通常是難以尋找的。本篇論文將呈現一個書評推薦系統，它從網際網路上的異質網站收集書評，自動執行品質評鑑，並且產出書評的優質排序。 書評推薦系統的核心－書評品質評估技術－是用來量測異質書評的品質，而品質預測結果將由專家來進行書評的優質排序之驗證。本論文所提出來的書評品質評估技術已被證明出可以產出優於其他方法的評估結果，像是商業網站上常用的按照書的評分排序，或是按照書評的發表日期排序等方法。此外，本篇論文也實作了在行動裝置上可以運行的書評推薦系統。在行動環境中，有兩個主要的問題，那就是行動裝置的運算能力較慢以及行動裝置所使用的無線網路頻寬不足。本篇論文採用分散式工作佇列以及JSON資料交換語言來解決行動環境中的兩個主要問題，在產出書評的優質排序更加地有效率。

One important issue at hand is readers usually have limited amounts of time to choose books to read from a market that is filled with similar types of books. In deciding which books to read, readers tend to depend on expert book reviews. However, it is very difficult to find quality book reviews. This dissertation presents a book review recommendation system that collects reviews from heterogeneous sources on the Internet, performs quality judgments automatically, and then generates top-ranked book reviews. The core of the book review recommendation system, review-evaluation techniques, was used to measure the quality of heterogeneous book reviews and those prediction results were then validated with a ranking list decided by experts. The proposed review-evaluation techniques was proved to produce better results than do other methods, such as ranking by rating score or by review date that are primarily used by commercial websites. Furthermore, a mobile version of the book review recommendation system was also implemented for handheld mobile devices. A distributed task queue and JSON data exchanging language was adopted to solve two major issues for mobile environment- slower computation ability and insufficient mobile network, for also generating quality book reviews efficiently.