On 2015 Dec 15, Diana Frame commented:

Interesting article and, as a researcher engaged in systematic reviews, I really appreciate the empirical testing done by the authors. I wonder whether the true benefit of machine learning algorithms will be not so much in performing the screening (what is often called "Level I" screening, on titles and abstracts), but in helping to plan search strategies. I could see using this tool at the very beginning of a project to find other non-obvious terms that describe the concepts to be covered in the review. Might be especially helpful as more and more full-text articles become available to churn through. Looking for a needle in a haystack can be challenging for human reviewers, but computers don't get tired.

On 2015 Dec 15, Diana Frame commented:

Interesting article and, as a researcher engaged in systematic reviews, I really appreciate the empirical testing done by the authors. I wonder whether the true benefit of machine learning algorithms will be not so much in performing the screening (what is often called "Level I" screening, on titles and abstracts), but in helping to plan search strategies. I could see using this tool at the very beginning of a project to find other non-obvious terms that describe the concepts to be covered in the review. Might be especially helpful as more and more full-text articles become available to churn through. Looking for a needle in a haystack can be challenging for human reviewers, but computers don't get tired.