設計與開發部落格輔助寫作系統

Abstract

寫作能力是一種綜合訓練，既是語言文字的訓練，同時也是思維能力的訓練。藉由寫作過程，可以訓練一個學生的思考、理解、推理、及創作等能力；同時亦可檢視該學生是否已理解本國語言文字並能加以靈活應用。在日常生活中，舉凡書信、公文、簡報、履歷表、及婚喪喜慶等各類應酬文書等，皆屬寫作的範疇。本研究結合人機互動與自然語言兩門領 域，提出並實作一套部落格文章輔助寫作系統，目的是幫助使用者在撰寫部落格 文章的過程中，若發生不知從何構思詞句下筆寫作時，本系統能夠適時適度的提 供文章寫作的概念和方向，激發使用者創作的泉源。由於部落格文章的主題相當廣泛，為提供使用者寫作的相關提示，我們使用網路上大量的部落格文本作為系統資料集，使參考提示的廣度足夠完善。另外， 我們亦提供人性化的操作介面，方便使用者撰寫文章。本系統以手寫板作為輸入介面，提供手寫辨識的功能，並給予常用詞彙的建議選項。系統自動分析使用者撰寫的句子，並萃取出句中的重點詞彙串列，組成查詢字串搜尋網路上相關的範例句子，最後產生文章接續可以撰寫的內容。在本研究中，我們提出一個架構於 HMM (Hidden Markov Model) 以及 PLSA (Probabilistic Latent Semantic Analysis) 之內容預測模型，該模型可運用於寫作預 測，而且也可被用於其它相關之自然語言研究上。我們預期經此計劃強化後的系統能進入實際運轉階段，也能為競爭日益激烈的海外中文學習市場提供有力的競爭工具。在第一期計畫中，本研究已經有初步成果，目前有一篇論文正於 IEEE Trans. On SMC – Part A 2nd revision，另外有一篇論文已經被 IEEE Trans. On SMC – Part C 接受將於近期刊出，以及兩篇期刊論文已經於 Expert Systems with Applications 刊出。

As the popularity of Internet grows rapidly, more and more people become used to sharing their ideas and experiences on their own blogs. Practically, blog writing is a good practice for people to use appropriate wordings to express ideas or feelings. Besides, people tend to use or to be inspired by some small text segments that have been appeared in other articles. Based on the above observations, this study proposes and implements a computer-aided system for blog article writing. The system assists users to compose articles, especially when they are not sure how to express their thoughts, not sure what to write or need some inspirations. Currently, the Web can be regarded as a big database, which contains a variety of blog articles. We have collected a huge amount of blog articles which will be used as system corpus in the project. The system adopts a handwriting pad as input interface. When a user is writing his or her blog, the system automatically analyzes current input sentence and extracts important words to generate a keyword list. With the keyword list, the system composes a query string to obtain candidate texts from the Internet. Finally, the proposed system ranks these texts for user’s reference, so that users can depict in the next sentence. In the first phase of the project, we have submitted one paper to IEEE Trans. on SMC – Part A. The paper is currently under the 2nd revision. In addition, one paper has been accepted to appear in IEEE Trans. On SMC – Part C and two other papers has been published in the journal of Expert Systems with Applications.