The Cleaned Repository of Annotated Personally Identifiable Information

Protecting student privacy is of paramount importance and has historically meant that that educational datasets are not released to the general community and instead shared among a small number of researchers working on specific projects. However, these datasets could provide significant value to the educational research community if they were made available and could help ensure replication studies of important educational research. Deidentifying the student data is, in some cases, sufficient to permit data sharing among researchers and even public release. However, most educational dataset are quite large, making deidentification extremely time-consuming and difficult. A solution is automated deidentification, but this is challenging for unstructured text data like that found in educational environments. This paper introduces a new open-source dataset called the Cleaned Repository of Annotated Personally Identifiable Information (CRAPII). CRAPII is designed to test and evaluate automated deidentification methods for educational data. The dataset comprises over 20,000 student essays that have been annotated for personally identifiable information (PII). Within the dataset, all occurrences of PII have been replaced with surrogate identifiers of the same type. The purpose of CRAPII is to promote the development of automated deidentification methods specifically designed for and tested on student writing. To further this goal, we are hosting a data science competition in which teams of data scientists compete to develop deidentification algorithms using CRAPII.