Project Number: 2019-128
Project Leader: Tom Coupe
Host Department: Economics and Finance
Project Title: Predicting Replications
Project outline: The goal of this project is to develop a program that automatically detects whether a scientific paper in economics is a replication of an earlier published paper. There are several archives of economics papers (Scopus, REPEC) and we would like to detect all replications in these databases so we can assess the rate of replication in economics. With the help of research assistants we have already manually extracted a list of replications but we now want to study whether it would be possible to automatize this process.
Hence, as part of this project, the researcher will need to write Python code to extract characteristics of papers [this can be from their abstracts, their Scopus descriptions or even the pdfs] and see whether we can use these characteristics to predict which paper is a likely replication. In addition, we want to use these paper characteristics to predict which replication is likely to be a successful replication.
Specific Requirements: advanced programming skills in Python, good econometrics skills.