Crowdnection: Connecting High-level Concepts with Historical Documents via Crowdsourcing

To form and test hypotheses and finally produce conclusions, people use existing schemas to search a pool of data for evidence. The quality of the search largely depends on the quality of connections between the schemas and the data. Making good connections between schemas and unprocessed data is challenging because it is time-consuming and may require expertise. Crowdsourcing provides a potential solution because with appropriate methods, humans are often more effective at synthesizing diverse information than automated techniques. This paper introduces Crowdnection, which leverages crowdsourcing methods to examine the effect of amount of context on performance in making connections between raw texts of historical textual documents and high-level concepts. The results suggest novices are able to help process information to provide meaningful insights, and indicate that there is an ideal amount of context facilitating the sensemaking process.