法令規則之萃取應用於醫院招標文件之檢核

Abstract

在國外，法律結合資訊科學的研究早已有相當豐碩的成果，但國內卻尚在起步階段，再加上中文的研究向來就比英文來得困難許多，因為中英文句子結構的不同，以及文件的多樣性，因此，我們的研究主要針對用詞簡單有規則性的中文法律條文，利用資訊萃取、自然語言處理的技術，來描述詞與詞之間的文法規則，我們想探討的是，對於這樣結構化的文件是否可透過程式的語意、句法剖析和關鍵詞擷取，將文字轉換成規則，擷取結構化資訊，建立儲存規則的規則資料庫。 本論文針對政府採購法萃取出來的規則應用於醫院採購招標文件之分析，擷取招標文件中的主要資訊與採購法萃取出來的規則做比對，檢查是否有與法令相牴觸，招標文件若有書寫上的錯誤，將會帶來很大的麻煩或嚴重的損失，因此，針對有誤的敘述，將提出該法條的內容警告或建議承辦人員，如此一來，就可以避免人為上的筆誤和疏失，減少機關單位的損失和與法規相違的可能性。在實驗結果中，針對醫院採購招標文件我們得到良好的結果，且系統擷取文件特定資訊和比對規則之檢核能力皆達90%以上。

In Taiwan, the application of information science in law is still in the beginning. Because Parsing Chinese documents is more difficult than English, and the structure of Chinese sentences is different from English. In this thesis, we propose an approach to describe the relationship between words in Chinese documents which was based on Information Extraction and Natural Language Processing technology. The sentence structure of Chinese legal clause is more pithy and regular, his structural document should be more feasible for structural information extraction by semantic or sentence structure parsing and keyword mapping. Finally we transform the legal information into rules and build the rule database. The rules extracted from Government Procurement Act were applied on automatic hospital procurement documents parsing. Extracting essential information in the procurement document and comparing with rules in database to check if there were contravene decree. If the documents have any mistake, it will bring big trouble or serious losses. Therefore, we can mark the error and give proper suggestion. Finally careless wrong statement can be avoided and possible losses of the organization and disobeying with the regulation can be minimized. In spite of the experiments show that our system can obtain well precision and high recall in hospital procurement documents drafting. This is a very limited application of text mining in Chinese documents. More work is needed in the future.