線上智慧型文章生成輔助系統

Abstract

本文以近千篇部落格文章做為系統產生的基底，使用中研院斷詞系統執行中 文斷詞工作，並加入 HowNet 輔助以及將 Google 搜尋當成語料庫，實做出一個 中文文章自動產生輔助系統。首先將三萬多篇部落格文章進行資訊和情感類的判 斷並加以分類，取出其中為情感類的文章，再加以做SPLR分析、整理，作為文章 產生的資料庫。系統會依使用者所期望之概念做詞義的延伸，再利用文章骨架作 關鍵字擴展，填充其內容而產生文章。產生的文章可利用Google即時擷取可替換 句子並做相似度分析，以讓使用者可替換所期望之填充句，已達輔助的功能。 由於在擷取文章資料庫時做過分類，故產生的出文章偏向日記類、部落格式 的文章，較貼近生活。而概念詞的輸入及擴展，以及相似替換句的擷取更能輔助 提供者產生期望之文章。期望藉由半自動、半輔助之方式，產生供使用者參考的 文章範本，以節省使用者在寫作時所花費的時間。

In this paper, we proposed and implemented an intelligent computer-aided article writing system, which makes use of about one thousand of essays that gathered from web blogs, and aided by using HowNet and Google search. First, we classify affective articles from 30000 articles that gathered from web blogs. Next, we reparse and reform those articles as our corpus by SPLR processing. The System will extend the concept entered by the user, and use the corpus to extend keyword list. Finally, the system generates the article based on the keyword list and content in the corpus. Meanwhile, the system allows the users to replace the sentences and the system will adopt Google as a corpus to provide some candidate sentences for user references. The system corpus was gathered from web blogs and that could add more variety to the content. The computer-aided writing system could help users organize content from concepts and reference the sentences written by other people. The first prototype system shows that it could assist the users to finish a blog article effectively.