從全文到表格：地方志職官志中職官資料之半自動擷取

本論文介紹一個針對中國地方志開發的電腦輔助全文標記系統，以幫助歷史學者及編輯人員從文本中有效率地蒐集「何人什麼時候任什麼官」的紀錄，將全文轉換成表格，作為大數據分析運算的基礎。中國地方志數量龐大，各地方志之職官志在描述「職官」時有大略一致的格式，然而在細節處差異仍大，難以運用電腦以統一的格式擷取各地方志的任官資料。本系統提供簡單易用的人機介面，內建擷取人名、官名、朝代、任職時間、籍貫、入仕等資訊的預設格式，讓使用者能簡單地按一個鍵套用，系統即辨識出文本中大多數的人名、官名、朝代、任職時間、籍貫、入仕等資訊，再供使用者校對。系統更提供使用者能因應文本格式差異、修改擷取格式的功能，以便從不同地方志中均能擷取到正確的任官資訊。

關鍵字

地方志；中國史；數位人文；資訊擷取；正規表達式

並列摘要

This paper introduces a semi-automatic text tagging interface to help historians efficiently garner posting records from the Chinese Local Gazetteers (difangzhi 地方志) with the format of ＂who, when, which posting.＂ By turning texts into tabular data forms, this interface aims to lay the basis for analyzing Chinese local gazetteers on a large scale. Although local gazetteers from various locations all follow a general pattern when recording posting data, they still differ in details due to their substantial amount. Therefore, it is unfeasible to ask programmers to extract the posting data using a onesize- fits-all computer program. This tagging interface, on the other hand, provides a simple user interface with built-in patterns to extract the subjects' names, posting titles, dynasties, posting times, basic addresses and entry methods. This allows users to tag most of data in the text quickly, which can then be proofread by them themselves in order to check the tagging result and to correct mistakes. The interface also enables users to adjust the extraction patterns according to each text in order to accurately extract posting data from local gazetteers with distinct patterns.