/* ---- Google Analytics Code Below */

Saturday, April 27, 2019

Automating Science Writing

How different is this from script writing mentioned previously?      Accurate summarization of text is another example of a challenge for AI type methods, similar in goal to finding a pattern in complex data.

Can Science Writing Be Automated? 
MIT News   by David L. Chandler

Massachusetts Institute of Technology researchers have developed artificial intelligence (AI) that can read scientific papers and produce a plain-English summary in a sentence or two. The team's neural network is based on vectors rotating in a multidimensional space, rather than on multiplication of matrices. The rotational unit of memory (RUM) system basically represents each word in the text by a vector in multidimensional space, with each subsequent word skewing this vector in some direction, represented by a theoretical space that can ultimately have thousands of dimensions. At the conclusion, the final vector or set of vectors is translated back into its corresponding string of words. Testing suggested RUM could be useful with natural language processing, so the researchers fed scientific papers through the network, generating a much more readable summary than a conventional neural network yielded  .... "

No comments: