Fastcase is a new legal research tool that is bringing cool lateral thinking to the traditional problem of serving up legal search results. The results are some highly impressive innovations both in the algorithms that sort and rank the results, as well as in how the results are presented.
Ed Walters, CEO of Fastcase, gave me an interview to describe their process.
1) You found that legal research has a three way tension between cases with the most citations, cases from the highest courts, and recency of the decision. How does your algorithm solve that problem?
For me, the problem with traditional approaches has been almost the opposite – there was no tension in results. They were always served up in a one-size-fits-all order, usually highest court first, then most recent case to oldest. That ordering makes sense sometimes, but more often it doesn’t. One size fits almost nobody.
So we did two things. First, starting way back in 2003, we gave users the tools to order results in the way that was most useful to them – more than a dozen options. We don’t resolve the tension, we let users resolve it for themselves, depending on the type of research they’re doing. We’re big believers in empowering people with smarter tools.
And some of these options were the first of their kind – like integrated citation analysis. We show the number of citations to each case in the search results – it’s like Shepardizing each case automatically, and no other service does that, even today. You can sort the most cited case to the top, and when the first case has been cited 4,000 times and the second has been cited 8 times, that tells you something very important that you wouldn’t know without Shepardizing every case in the list.
Second, we resolved the tension by building an Interactive Timeline view of results – a 4D map of all of the results plotted over time. You can see at a glance how the results plot over time, which ones are the most relevant, decided by the highest court, most cited anywhere else.
You have to see it to believe it – when you see search results represented graphically, you can see how they score on each metric instantly [emphasis added] – so the tension between the different measures is resolved, and the right answer literally jumps off the page.
2) The idea of using citations the same way search engines use links – as a key ranking factor – is simple, yet the devil with any programming project is always in the details. What obstacles have you overcome so far in developing the software?
One of the biggest challenges was to give context and meaning to the citation analysis. When you run a search about a standard of review, you might get a case like Brown v. Board of Education in the search results, and if you sort by how often cases are cited, Brown would be at the top of the list. But Brown is a desegregation case, not a case about the standard of review.
So we built a tool that shows how many times each case has been cited only within the set of super-relevant search results. “Cited Within” shows with a click which case is most cited within your specific topic (and on your point of law). Again, this kind of citation analysis is only available on Fastcase.
The super-relevant set of results is basically the search results returned by the user’s search. When the user runs a keyword search, he or she gets back a set of search results (say, for example, 408 cases). We show how many times those 408 cases are cited anywhere else in the database in general – but some of those cases may be frequently cited for a different point of law. (Therefore) We also show how many times they’ve been cited by the other 407 search results – and most often that’s what the user cares about.
It’s a heuristic – that tells you not only which case has been cited most, but also which case has been cited most by the other cases you care about. And it’s custom-built into each search.
3) How do your visual search results make life easier for researchers?
Some people are just visual learners – for them, long lists of text results don’t tell the same story that an elegant data visualization does. Instead of viewing 20 or 100 results at a time, you can see all of the results on a single map that you can customize, zoom, and craft.
It’s beautiful – it’s also a ruthlessly efficient way to find what’s most important. For a visual learner, there’s really no way to be happy with text-based search results after viewing results on a visualized Interactive Timeline map.
4) Free legal research sites like CanLII give you the option to sort by most-cited or most recent. How does your search engine provide greater value (imagining you offered it in Canada)?
I love CanLII and Cornell’s Legal Information Institute – they’re terrific services and leaders in the field. We share a common goal of democratizing the law. But where the goal of an LII is to make the law accessible to everyone, at Fastcase we’re really trying to build the smartest, most powerful tools for lawyers (although there are nonlawyer subscribers, the vast majority of our more than 500,000 paid subscribers are lawyers).
So we focus on things that are more important to lawyers, like our Authority Check that shows later-citing cases, powerful search syntax that works whether you use traditional Boolean-style keywords or free-form Google-like searches. We focus on things like dual-column printing in PDF, Rich Text Format, and Word *.doc format, batch printing for multiple documents at once, customizable search histories and libraries for saving documents, and of course, data visualization and integrated citation analysis.
I’m really excited about Forecite, a feature we just rolled out, as an example of smarter tools for research. When you run a search, Forecite automatically reads all the cases in a search result and builds a table of all the cases that are cited. If a case is frequently cited by the search results – but doesn’t appear in the search results – Forecite automatically suggests it.
Forecite only suggests additional resources where the citation analysis says there’s something important that you’d otherwise miss – but you’d be surprised at how often that happens. If the seminal case is outside your jurisdiction, or uses older terminology than your more modern keyword search, Forecite acts as a safety net and delivers the seminal case you’d otherwise miss. It’s like magic – you can see more at www.fastcase.com/forecite.
5) When you started the company, how did you keep up your motivation? What were the hardest parts of building the business?
The biggest challenge has been assembling the legal information – from thousands of sources in many different formats. That’s why we’ve worked hard in the Law.gov effort, working in 2008 with Public.Resource.org to create the world’s first public archive of U.S. Supreme Court cases and cases from the Circuit Courts of Appeal. In 2011, we’ll launch the Report of Current Opinions (RECOP) with Public Resource, which is a weekly collection of cases decided in federal and state Supreme Courts and Courts of Appeal.
Getting access to public domain law shouldn’t be the hard part of this process – everyone should have that. Publishers should be competing on who can build the smartest tools, deliver the best service, and at the highest value. That’s a competition that we’ve geared up to win, and we’re proud to work with the Law.gov effort to make that a reality.
If you liked this interview, check out other posts in the interview, legal or algo categories. Also, add my RSS feed to your reader so you can stay up to date with the latest posts as I publish them, or get my latest posts by email.Algorithms, Interviews, Law