Meet Sherlock, a concept extractor like no other

Continue slides....
Over loaded with text ?
Under informed by numbers ?
You've come to the right place!

Directories gave you documents by category

Search engines then gave you a lot more 

Now Sherlock delivers key points, connections, and evidence automatically

and then delivers the key points and connections, no bias added.

Sherlock grinds documents into yes/no questions about the words contained

removes the junk, tests for associations,

e.g. The example below says that the first document ( row ) contains the first word in the dictionary ( column ), only the second document does not.

 

Example: Sherlock's first paid gig.  Wall St. hedge fund manager wanted the US election covered. Here are 500 tweets from early August to slice and dice. Browser full version available here .

See below the sort of output that Sherlock delivered 24/7, real time.

One cluster looks backwards, seven nodes

Another uses "AGAIN", looks forward, four nodes.

and delivers the key points and connections, no bias added.

Sherlock grinds spreadsheets into yes/no questions about the numbers contained

removes the junk, tests for associations,

e.g. Imagine a spreadsheet of numbers in which you want to find increases that happen together.

 

The example below says that the first number ( column )  increases except in case two ( row ).

 

Numbers, changes, and rules

Proof of Concept : Row 5

EUR/USD up whenever EUR/GBP up AND GBP/USD up

( 229 times , EUR/USD is up 618 times in total so other rules may apply )

Sherlock reduces overload by automatically exposing the connections between any symbols, uniquely using logic , rather than statistics, to justify and grade the patterns. So Sherlock can associate words in any language with sounds, pictures, or any other media.

 

Logic also drives results that are both useful, because they are expressed as clear rules, and trustworthy, because they come with a full audit trail back to the evidence.

Testimonials & More below....

If you think this is for you, as user or backer, then you should reach out now to start the conversation. Otherwise follow the trail below , and get convinced!

Chris Painter and his Meme Machines’ Sherlock system for deep understanding of text is a revelation.


In years of working on machine learning, the tough nut to crack has been methodologically sound textual analysis that  integrates with other structured forms of predictive analytics.
                   
We found Sherlock combined rigour with elegance to power the deep textual insights we needed in finance and technology.
                      

Professor Michael Mainelli, Executive Chairman Z/Yen

Z/Yen is the City of London's leading commercial think-tank, founded in 1994 to promote societal advance through better finance and technology.

 

 

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.

Vinay Gupta, Associate Fellow at the Institute for Security and Resilience Studies, University College, London University

The ease by which concept  extraction can switch from English to Arabic shows that Sherlock can work on any "text". Sequence doesn't matter, Arabic reads from  right to left after all, what matters is proximity of words or  attributes to each other.

Doctor James Little, Department of Mathematics, Çankaya University, Ankara, Turkey

Mining the Simple English Wikipedia , 3000 at a time, working towards a full atlas  of world knowledge, a full Wikipedia theme meta-index.

Output after first 3000 documents mined; as hoped, two clear clusters, Geography and Zoology. Another Proof of Concept.

Help Sherlock reach its potential!

Have your say in the Ideas Box

Co-operation? Investment? Need more explanation?