Behind the Scenes at Prophy

How Our AI Streamlines the Peer Review Process

Trends for now:

• Peer review is a core element of the scholarly publishing process, but the infrastructure supporting it is embedded within publisher systems and workflows. 
• There are some open peer review platforms attempting to improve peer-review, however adoption is slow..
• AI and machine learning is being increasingly employed to bring efficiencies to the peer review process and improve quality.
• Research integrity is called out as an area needing more investment. In particular developing secure researcher identities to help guard against fraud and misconduct in peer review.
• Peer review still remains a core validation mechanism for much of scholarly publishing. So improving the efficiency, quality, and integrity of peer review through shared infrastructure innovation could have real value.

At Prophy, our mission is to enhance scholarly peer review using the power of artificial intelligence. We’ve built an advanced AI system that deeply understands the concepts within academic papers across all scientific disciplines. Our knowledge base includes over 160 million papers, 70 million authors, 70,000 journals, 100,000 institutions from 224 countries, and 140,000 scientific concepts! In this blog, we’ll explain how our technology works to radically streamline the peer review process.

Our AI performs the following key steps:

• Natural Language Processing: We extract key conceptual phrases from the full text of manuscripts, avoids reliance on ambiguous author keywords.
• Concept Vectorization: We embeds concepts mathematically into a high-dimensional vector space, encodes semantic meaning.
• Manuscript Vectorization: We aggregates concept vectors into a single vector “fingerprint” summarizing the paper’s content.
• Similarity Matching: This enables us to rapidly compare manuscript fingerprints to our database of 70M+ researcher fingerprints to find the most similar candidates.
• Ranking and Filtering: We then ranks the most similar candidates using additional criteria like diversity, availability, conflicts of interest, etc. 
• Recommendation: Finally we suggests the best reviewers to editors within seconds using integrated APIs.

The magic happens when our AI matches the manuscript fingerprint to our database of papers and researchers. It rapidly finds the most conceptually similar candidates to suggest as peer reviewers.

The top matches are further ranked using additional criteria like reviewer diversity, availability, conflicts of interest, and journal fit. Only the best matches are recommended to editors. 

Our API integration allows this AI-powered suggestion process to happen seamlessly in existing editorial workflows. Editors get relevant reviewer recommendations in seconds!

By leveraging AI to deeply understand papers based on their concepts, Prophy provides unbiased, relevant suggestions even for interdisciplinary work. As our knowledge base grows, so does the power of our matching. 

The automation and sophistication of our concept modeling is what allows Prophy to massively improve the efficiency and quality of peer review. We’re excited to continue developing cutting-edge AI to further enhance this critical part of science.