台灣的國會選舉中，一向規定個別立法委員候選人需提出政見，並刊載在選舉公報上。雖然目前坊間主流意見多認為政見，尤其登之於公報上的政見，一般選民根本不看，與候選人當選與否亦應是關係不大，所以相關的研究並不多。但本研究認為公報上的政見，既是候選人正式的選前承諾，不應積非成是，在研究上就理所當然予以忽視；反而應藉由探討這些承諾與選後國會中表現的聯結情形，來強化選民對代表的「課責性」；也希望因而能促進各方對公報上「政見」的重視，形成民主政治的良性循環。為了達到上述目標，本文針對我國第六、第七屆立委選舉政見，建構「第六及第七屆立委政見資料庫」，希望藉此資料庫的建置，未來探討如立委選前政見與選後立法表現等相關研究議題。本文主要說明此資料庫的建構過程，包括所用的人工登錄方式和資料採礦的資訊技術，特別是新進發展的多重標籤分類方式；本文並企圖透過這兩種方式的比對，未來能引進資訊技術，對每一次選舉的公報政見，能做相當程度的自動化分類處理。 In Taiwan's legislative elections, candidates running for office are required to submit campaign promises to be published in the Electoral Bulletin. This research considers that campaign promises published in Electoral Bulletins are the official commitments of the candidates and should not be disregarded as insignificant and unworthy of research. Moreover, the connection between the candidates' campaign promises and their performances after being elected should be studied in order to monitor their ”accountability” to voters. It is our hope that more emphasis is placed on ”campaign promises” printed in Electoral Bulletins, which would help to create a positive influence on the creation of a democracy.In order to achieve the above stated goal, this paper has established the ”Sixth and Seventh Legislative Campaign Promise Database” based on campaign promises by candidates during Taiwan's sixth and seventh term legislative elections. Hopefully this database would facilitate the research on issues related to the connection between campaign promises and the legislators' on-job performance after being elected.This paper will mainly cover the establishment process of this database, including the manual coding method and the latest information technology used for data mining, namely the multi-label classification method. This paper also attempts to compare and contrast these two methods and hopes to bring in the most advanced information technology to conduct a certain level of automated classification for campaign promises printed in Electoral Bulletins in every election.