檔案主題新聞之數位人文研究

我國檔案管理局對於檔案公開應用，不遺餘力的推動，但社會大眾對於檔案開放應用的瞭解仍然不足。有鑑於新聞報導為社會大眾瞭解時事議題的重要媒介，政府機關可透過新聞資訊的傳播，讓社會大眾對於政策作為有所瞭解，且能理解施政的表現。如能藉由資料探勘方式，預測新聞報導中呈現的主題與情感意涵，應可提供決策者發布新聞及營造良好公共關係的參考。本研究藉由文獻回顧瞭解中文斷詞以及情感分析的資源，以資料探勘的方式，對檔案主題新聞文本進行斷詞處理，並透過詞性標註與詞頻計算結果，找出檔案新聞文本報導主題之詞彙。再進一步比對CVAW 4.0情感詞典，找出具有情感意涵之詞彙，透過情感分類模型的建立，預測434篇檔案主題新聞文本之正面、中性、負面傾向分類，以瞭解檔案新聞文本中報導主題之情感意涵。研究結果發現，透過詞性中的專有名詞標註以及高詞頻的篩選，找出檔案主題新聞所報導的主題詞彙有「二二八事件」、「白色恐怖」、「美麗島事件」、「陳水扁」、「施明德」、「陳豐義」、「蔣介石與蔣中正」、「尤美女」、「雷震」、「希拉蕊」等詞彙，透過瀏覽包含這些高頻詞彙之報導內容，並分析其情感意涵，發現此類報導多呈現負面情感意涵，根據研究結果建議檔案管理機關對於民眾關注之檔案徵集、管理及應用問題，應加強作業改善和檔案素養教育推廣，以期符合社會大眾對於檔案使用的期待。

關鍵字

檔案新聞；數位人文；資料探勘； CKIP中文斷詞系統

並列摘要

The National Archives Administration has promoted the application of archives for many years, but the public still has insufficient knowledge about the archives. Government agencies can use the dissemination of news to make the public understand the policy and realize the performance of governance. In addition, with the development of information technology, through the way of data mining can simplify manual interpretation method to predict the themes and emotion of the news that can be provide references for policy decisions in the future. This research used literature review to understand the concepts of Chinese word segmentation and sentiment analysis, and uses the data mining method to process word segmentation of archive news texts. According to the results of characteristic tagging and word frequency calculation, find the vocabulary of the topics in archive news text. Further, the CVAW 4.0 sentiment dictionary was used to find words with emotional meaning, and through the establishment of the sentiment classification model, predict the classification of positive, neutral, and negative tendencies of 434 archive news texts, in order to understand the emotional meaning of archive news texts. The results of the study found that through the use of characteristic tagging and high-frequency filtering, the topic vocabularies of the archive news text were "February 28", "White Terror", "Beautiful Island Incident", "Chen, Shui-bian", "Shih, Ming-teh", " Chen, Feng-yi", "Jiang, Kai-shek and Jiang, Zhongzheng", " Yu, Mei-nu ", "Lei Zhen", "Hilary" et al. This research analyzed its emotional connotation and found that such texts often present negative emotional connotations. Based on the research results, the suggestions were proposed for the archival agencies should strengthen the improvement of work process and encourage the archival literacy education for the public in order to conform to the public's expectations for the use of archives.