The second project idea is about the application of deep learning-based text classification models within the context of Automation of Systematic Literature Reviews (SLR). We want to determine which articles are relevant or non-relevant (binary classification).

Ive got data from Cohen et al. (2006) from his personal website: https://dmice.ohsu.edu/cohenaa/systematic-drug-class-review-data.html and from Howard et al. (2016) from Springer, additional files appendix 1 and 2: https://link.springer.com/article/10.1186/s13643-016-0263-z As they are lists of PubMed IDs, you need to create a request for every single ID, which indeed takes a lot of time. Therefore, I share with you three datasets including all metadata. You can find these files under Project Related Files directory in this itslearning course page. You can see two datasets with excluded/included labels. Abstract and title of each article exist in a row of this file together with the label.

To get more information about this SLR automation idea, please see the following article: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1447545/ Reducing Workload in Systematic Review Preparation Using Automated Citation Classification A.M. Cohen, MD, MS, W.R. Hersh, MD, K. Peterson, MS, and Po-Yin Yen, MS Here you can also see some discussions regarding the implementation, esp. optimization issues. https://stackoverflow.com/questions/64264583/custom-keras-metrics-class-metric-at-a-certain-recall-value if you need anything let me know