Improving the Efficiency of Militarized Interstate Dispute Data Collection using Automated Textual Analysis

军事化国家间争端（MID）数据集是国际冲突研究中最突出和最常用的数据集。更新这个数据集已经变得既费力又昂贵，其结果是一个延迟更新的周期，并且当更新发生时成本不断上升。为了持续地维护数据，需要一种更流畅、更小、更有效和更便宜的方法。该项目建议使用自动文本分析算法来识别包含可编码的军事州际行动的新闻故事，这是该过程中最昂贵和劳动密集型的部分。目标是产生一个搜索过程，与以前的人类数据收集非常相似，允许在将来更及时、更定期地更新MID数据集。该项目将修改搜索引擎TABARI（文本分析增强替换指令），该搜索引擎是堪萨斯事件数据项目（KEDS）的一部分，用于识别包含MID项目感兴趣的事件信息的新闻报道。该项目将通过对已经编码为2001年的新闻报道进行试验来实现这一目标。我们想要改进TABARI，这样它就可以在不产生大量不相关报道的情况下找到人类编码人员之前识别的新闻故事。人类编码员将阅读TABARI确定的新闻报道，并验证他们是否有相关材料或确定所报道的事件不构成军事事件。我们希望改进搜索过程，以便在下一次广泛的MID数据收集和随后的扩展中，将新例程应用于收集信息的任务。该项目的更广泛意义在于，成功完成将为2002年至2008年期间MID数据集的更新以及此后几乎连续的更新提供基础。由于MID数据被学术界广泛使用，频繁的更新将有利于研究人员解决广泛的研究问题，包括但不限于民主和平、军备竞赛、联盟、冲突管理、转移使用武力、政治地理和领土争端、危机升级、区域安全、贸易和相互依存、国际规范、集体安全、军事战略、巴勒斯坦态度、干预、以及早期预警系统。特别是，将数据集扩展到这个新时期，将允许对冷战后和9/11后世界中有关国际冲突的问题进行更有力的审查。此外，该项目将说明这种类型的自动编码在学者广泛使用的社会科学中的其他数据集的可行性。

The Militarized Interstate Disputes (MID) dataset is the most prominent and frequently used collection in the study of international conflict. Updating this dataset has become both labor-intensive and prohibitively expensive, with the result being a cycle of delayed updates and escalating costs when updates do occur. To maintain the data on a continuing basis requires a smoother, smaller, more efficient, and less expensive methodology. This project proposes to use automated text analysis algorithms to identify news stories that contain codable militarized interstate actions, which are the most costly and labor-intensive part of the process. The goal is to produce a search process that closely emulates previous human data collection, allowing for more timely and regular updates of the MID dataset in the future. This project will modify a search-engine - TABARI (Textual Analysis By Augmented Replacement Instructions) - that was developed as part of the Kansas Event Data Project (KEDS) to identify news reports that contain information about events of interest to the MID project. The project will do this by experimenting on news reports that have already been coded for the year 2001. We want to refine TABARI so that it can find the news stories that human coders previously identified without producing large numbers of irrelevant reports. Human coders will read the news stories identified by TABARI and verify that they have relevant material or determine that the events reported do not constitute militarized incidents. We expect to refine search procedures so that the new routine can be applied to the task of gathering information in the next broad MID data collection and in subsequent expansions.The broader significance of this project is that successful completion will provide the basis for an update of the MID dataset for the period 2002 through 2008 and nearly continuous updates thereafter. Because the MID data is so widely used by the scholarly community, frequent updates will benefit researchers addressing a wide range of research questions, including but not limited to the democratic peace, arms races, alliances, conflict management, diversionary uses of force, political geography and territorial disputes, crisis escalation, regional security, trade and interdependence, international norms, collective security, military strategy, Palestinian attitudes, intervention, and early warning systems. In particular, expansion of the data set to this new period will allow for stronger examination of issues regarding international conflict in the post-Cold War and post-9/11 world. Moreover, this project will illustrate the feasibility of this type of automated coding for other datasets in the social sciences that are used widely by scholars.