How Will He Solve the Case?
Sherlock is the world’s first AI-powered digital document bloodhound. In milliseconds, Sherlock analyzes documents you have seen and then finds other relevant ones to further your investigation.
This page lets you see Sherlock in action, solving some of his most famous cases. We start with several introductory cases and then move on to some more difficult ones. Ultimately the goal is to demonstrate the power and flexibility of our Search 2.0 platform and to have a little fun and A.C. Doyle’s expense.
Sherlock is a machine learning algorithm, that learns from your training. As you send documents to Sherlock, you are giving him signals about what you are looking for. With Pandora, it is about musical tastes. With Sherlock it is about relevant document content.
This is an all new way to search, which is why we call it Search 2.0. As you will quickly see, keyword search alone (aka Search 1.0) is quite limited in its capabilities. Because of inherent limitations in language, keyword search tends to bring back too few good documents and way too many bad ones. You pay the price in time and frustration. And “time is money,” as they say.
An Introduction to Sherlock®
We made these videos for fun and to give you an idea of Sherlock’s speed and power. Our site contains about 1.9 million emails created during Jeb Bush’s two terms as governor of Florida in the early 2000s. Governor Bush made these documents available to the public and a portion were used at the NIST Text Retrieval Conferences (TREC) in prior years.
We found even more of the Bush emails and thought they would be useful to show Sherlock’s ability to find helpful documents on a variety of real topics. Each came in the form of a document request which we used to start the process.
Watch as Sherlock gets on the trail. Go get ’em Sherlock.
Sherlock and the Case of the Protected Manatees
In this introduction to Sherlock, we start with one document relating to efforts to protect manatees in Florida. We send it to Sherlock which then analyzes approximately 1.9 million documents on the site to find more. Watch how quickly Sherlock finds helpful documents.
Keyword search could be difficult here because we don’t want documents about Manatee County Florida, which outnumber those relating to the animal we want to protect. Can Sherlock distinguish between the two? Let’s find out.
Some of Sherlock's more difficult cases.
Sherlock and the Case of the Misspelled “Havanna Club”
This time our searcher made a mistake. She misspelled Havana Club adding an extra N. The search found several good emails where the person sending them didn’t realize that Havana only has one N. It turns out there were a lot more documents on this subject and even more that didn’t have the words Havana Club in them.
Can Sherlock help find other relevant docs even with a bad keyword search? And can Sherlock move beyond the phrase “Havana Club”?
Sherlock and the Case of the Florida Ticket Scalpers
Our mission today is to track down emails discussing Florida’s plan to legalize ticket scalping. There was a bill before the Florida legislature that would soon head toward Governor Bush’s desk. Some supported it, others urged a veto.
In this case, Sherlock got confused because an early document had lots of irrelevant HTML links in it. After a few “Thumbs Down” signals, Sherlock got right back on the track.
Sherlock and the Case of Florida’s Library Closure
In testing Merlin Integrated Search, we stumbled on a clue suggesting that a move was afoot to eliminate Florida’s state library system, turning it over to private hands. Would Governor Jeb Bush take such a step and make Florida the “laughingstock of library circles?”
Starting with a pretty weak document, watch how quickly Sherlock gets on the case.
Sherlock and the Case of the Looming Climate Crisis
Sherlock’s mission today is to find documents relating to the debate over global warming and climate change and to help determine Governor Bush’s position on the subject. The problem is that many of the documents Sherlock finds are similar in content and language.
Traditional document reviewers have to go document by document to find relevant information. Sherlock has a better way. Watch has he brings back 50 documents at a time, clusters them by similar content and allows reviewers to intelligently tag clusters with a single click. Can we review 50 documents in 5 rather than 50 minutes?
Watch this episode and see for yourself.
Integrating advanced keyword and algorithmic search
Sherlock and the Case of the Voting Felons
In this episode we show a new way to search for documents. We received a document request regarding “Felon Disenfranchisement.” Rather than build a keyword search, we simply pasted the full request into Sherlock Integrated Search and brought back results in relevance order.
We then selected the top results and sent them to Sherlock. Immediately we were on the trail to find more documents about restoring voting rights to felons.
Sherlock and the Case of Florida’s Disappearing Water.
This is another example of Search 2.0 and our radical new approach to integrating keyword and algorithmic search.
This document request was about bottled water and a concern that companies like Perrier and Nestle were draining Florida’s natural springs. Instead of crafting a complex keyword search, we simply pasted the request verbatim and let Sherlock go. This is another good example of Search 2.0 in action.