中文情書自動產生系統

Abstract

本文以數百篇情書做為系統產生的基底，配合中央研究院的平衡語料庫，實做出一個中文情書自動產生系統。首先將情書語料庫和平衡語料庫的文章進行分解、重新整理，再利用處理過後產生的文字片段組合出一篇新的情書文章。將文章中最有意義的關鍵字取出，並以隱藏方式延展擴增成一關鍵字串列，再用填充詞的方法將關鍵字予以串連，以成為一篇全新的情書文章。 產生出來的情書文章無論在取材、句型、創新、文意上都有不錯的效果。本系統仍在發展階段，所以還有改善的空間，但現階段已表達出一般人在寫作時的進行方式與情況。

In this paper, we proposed and implemented an automatic system producing the Chinese love letter ,which makes use of hundreds of essays about love letter topic and corporate with Sinica Corpus to produce new love letter. First, rearrange the corpus based on the love letter collection and corpus from Sinica Corpus and then combine the material to produce new love letter. The most meaningful keywords in the article will be retrieved ,and extend the method to hide amplified into a serial keyword, and then filled with the word method will be the keyword link to become a new love letter article. As our experience with a first simple prototype has shown that this approach could produce the love letters. There is a good result no matter materials, on sentence pattern, innovation, gentle purpose in the love letter article produced out. This system is still during the course of developing, so there is improved space, but has already expressed carrying on the track type and situation while writing of common people at the present stage.