TL;DR AI Summarises Research

The single most significant starting point for a researcher in any field is the Literature Review. This is where the researcher reads through large troves of research papers in order to absorb and understand the current state of research in their chosen field.

Wouldn't it be wonderful if someone (or something) could summarise these papers so that only relevant research could be chosen in order to do a deep-dive? 

Researchers at the Allen Institute for Artificial Intelligence have developed just such an AI model. It summarises text from scientific papers and presents them in a few sentences in the form of TL;DR (Too Long; Didn't Read). 

The AI model takes the most important parts from the Abstract, Introduction, and Conclusion section of a research paper to form the summary.

Researchers first “pre-trained” the model on the English language. Then they created a SciTLDR data set of over 5,400 summaries of computer science papers. It was further trained on more than 20,000 titles of research papers to reduce dependency on domain knowledge while writing a synopsis.

The trained model was able to summarise documents over 5,000 words in just 21 words on an average — that’s a compression ratio of 238. The researchers now want to expand this model to papers in fields other than computer science.

There's still further work to be done in this area such as using specific or unique properties of papers in a certain domain or field, including the full-text during training rather than just the Abstract-Introduction-Conclusion sections, and also take into consideration that the idea of TL;DR can differ among academic disciplines.

Check out the Semantic Scholar search engine where you can see TL;DR AI summaries for Computer Science research papers.

Plus here is access to the research paper on this advancement.