Artigo Acesso aberto Revisado por pares

Crowdsourcing the Measurement of Interstate Conflict

2016; Public Library of Science; Volume: 11; Issue: 6 Linguagem: Inglês

10.1371/journal.pone.0156527

ISSN

1932-6203

Autores

Vito D’Orazio, Michael Kenwick, Matthew Warren-James, Glenn Palmer, David Reitter,

Tópico(s)

Sports Analytics and Performance

Resumo

Much of the data used to measure conflict is extracted from news reports. This is typically accomplished using either expert coders to quantify the relevant information or machine coders to automatically extract data from documents. Although expert coding is costly, it produces quality data. Machine coding is fast and inexpensive, but the data are noisy. To diminish the severity of this tradeoff, we introduce a method for analyzing news documents that uses crowdsourcing, supplemented with computational approaches. The new method is tested on documents about Militarized Interstate Disputes, and its accuracy ranges between about 68 and 76 percent. This is shown to be a considerable improvement over automated coding, and to cost less and be much faster than expert coding.

Referência(s)