"Sherlock's correlation engine seems to me like it could become as common as search button functionality is today: an automated way to make sense of complex document collections"
"We found Sherlock combined rigour with elegance to power the deep textual insights we needed in finance and technology"
Emeritus Gresham Professor of Commerce at Gresham College in London, founder of the Long Finance initiative, he was elected Sheriff of the City of London on 24 June 2019.
Produced by Morgan State University, this diagram describes the benefits of the approach. Their website has a very readable tutorial available, in which they spell out guidelines for the manual production of maps. Sherlock follows them, and is, as far as we know, the first autonomous process on the planet to do so.
It may sound odd to think of an idea as something that can be found in text, but that is how we store and transfer ideas. We package them in sequential nested chunks that make them easier to digest. Once the packaging is removed the patterns emerge, as words that are frequently seen together to denote the underlying idea.
The word "idea" is derived from old Greek that means "to have seen". If you want to hear how noticing coincidences formed the original notion of 'idea', and how it remains central to Cognitive Science, check out this podcast from Oxford's Bettany Hughes
Sherlock finds the common ideas that bind groups together, and needs no further information than the words written.
The process is easily explained, and fully traceable, because the answers are in the sources, verbatim.
It doesn't need, use, or store private data. Privacy is baked into the design.
No more retraining, no more parameter tweaking. If an earthquake occurs, or a new political sensation invades the press, new names will baffle supervised text-mining techniques.
Sherlock works from a list of what to ignore, so new stuff gets handled automatically, just so long as it is frequent.
Sherlock tries to be as autonomous as possible, requiring just two parameters, how robust the patterns need to be, and how many words will be enough.
For example, the UK Brexit survey takes 75% pattern confidence, and 20 words
© Copyright 2014-2020 Christopher Painter- All rights reserved.