Development of Machine-Coded Event Data Techniques for the Analysis of Political Behavior

点击翻译按钮获取中文摘要

9410023 Schrodt This project extends and demonstrates the use of software for the automated coding of political event data. Event data are generated by coding news reports for specific types of political interactions: meetings, agreements, threats, military engagements, etc. These data can then be used to test hypotheses about political behavior. Historically, human coders created event data. Earlier work supported in part by the NSF's Data Development in International Relations project demonstrated, however, that machine coding is capable of assigning event data codes to wire service reports of political events with a reliability at least as great as that achieved by human coders. The initial machine-coding system was tested by generating an 11- year event data set based on Reuters news service reports for the Middle East. These data produced statistical results almost indistinguishable from those produced from human-coded data. The data set and the machine-coding program have been available to political science researchers for the past two years; the system runs on conventional microcomputers. The current project will extend this research in three ways. First, the project will develop an integrated data acquisition and coding system to code event data in real time. These data will be made available on an Internet server for use by other researchers. This system will provide a significant improvement over existing human-coded event data, which typically are not available until years after events have occurred. Second, the project will enhance the machine coding program in several ways that have been suggested by research experience with the program. Finally, the project will use the machine-coded data to study two cases of complex political behavior: the Israeli-Palestinian conflict and the ECOWAS international subsystem in West Africa. These two cases are designed to demonstrate the use of the full capabilities of machine coding, particula rly the facility for re- coding texts to detect political activities that may have been missed in the standard event coding systems. In addition to developing techniques for event data analysis, these studies will contribute to understanding the interactions between the international and domestic political systems.