Wednesday, January 23, 2013

WestlawNext vs. LexisAdvance vs. Google Scholar

What's the best search tool for finding the most relevant case law on some legal topic? WestlawNext and LexisAdvance both claim to allow searchers to quickly locate the most relevant content with simplified Google-like search boxes, and Google Scholar also searches cases. So I decided to compare the relevance sorting for the three products using some basic searches for patent law topics (i.e., without using tools such as SY,DI() or ATLEAST). In general, I found that WestlawNext did the best job pulling up the most relevant cases—but if you really know nothing about a topic, you are still probably better off starting with treatises and other secondary sources (e.g., for patents, start with Lexis's Chisum treatise). (For more on doing patent research, see my page on patent references.)

How we search for legal information online is an important and under-discussed topic. Law students and lawyers spend a vast amount of time doing (and summarizing) legal research: student RAs do research for law professors; law clerks do research for judges; law firm associates do research for partners. But how often do professors, judges, or law firm partners ask where the information in a research memo came from? They might ask some general questions (did you look for Supreme Court cases? did you look in Chisum?), but I suspect questions about specific search queries are rare. And explicit discussions about search—i.e., more than the typical unhelpful Westlaw and Lexis training sessions—are probably more important as students increasingly expect these search tools to locate the most relevant cases as easily as Google locates the products they want to buy.

Searching for the Most Relevant Cases on Patent Law Topics

As discussed in detail at the bottom of this post, how you construct searches matters. I did a variety of searches on each of the following topics:
  • Written Description: WestlawNext's first result is Ariad v. Eli Lilly, the 2010 en banc Federal Circuit decision that reaffirmed that the written description requirement is separate from enablement, which is probably the first case I would want to look to (and which occupies a substantial portion of Chisum § 7.04, the section on the written description requirement). LexisAdvance and Google Scholar seem thrown off by highly cited cases that use the phrase "written description" but that are not about the written description requirement: depending on how the search is constructed, Lexis tends to start with Markman (either the Federal Circuit or the Supreme Court opinion), with Ariad at #5, #9, or #132. Google starts with Phillips v. AWH and has Ariad at #6 or #17. (Relevance is somewhat subjective, but I suspect most people searching for "written description" aren't trying to find claim construction cases.)
  • Utility: Westlaw starts with In re '318 Patent Infringement Litigation and In re Fisher, two recent on-point Federal Circuit cases. Google and Lexis start with Graham v. John Deere (a Supreme Court case on obviousness, not utility, though it uses the word "utility").
  • Inherent Anticipation: Westlaw starts with Schering v. Geneva and SmithKline Beecham v. Apotex, two recent on-point Federal Circuit cases (which are discussed thoroughly in Chisum § 3.03 on inherent anticipation). Lexis puts these at #4 and #5 and starts with Graham, the Supreme Court obviousness case. Google starts with Schering and then has Continental Can, an older Federal Circuit case which does discuss inherent anticipation, but which only merits one sentence in Chisum (noting that it was distinguished by Schering). (Or, if the Lexis search is for [ patent utility ] rather than [ patent and utility ], the first result is a case on rates charged by a utility company, where a separate opinion refers to something that "is patently a wholly gratuitous assertion as to Constitutional law.")
  • Divided Infringement: Westlaw and Lexis both put Akamai (the August 2012 en banc Federal Circuit decision) at #5. Results #1-4 are more relevant on Westlaw. Akamai is way down in the Google results.
  • Inequitable Conduct: Westlaw starts with Therasense (the 2011 en banc Federal Circuit decision), which is #10 on Google and #50 on Lexis. But Google and Lexis at least start with Federal Circuit inequitable conduct decisions—Google's #1 result is the 1988 en banc decision in Kingsdown; Lexis's is Molins, a 1995 panel decision; both are discussed in Chisum § 19.03).
Of course, one can't generalize too much from a few examples, and the results might be very different for non-patent topics. If you see very different sort-by-relevance results for WestlawNext, LexisAdvance, and Google Scholar for some other topic, please leave a note in the comments! Also note that I haven't compared classic Westlaw and Lexis—you can sort cases by relevance in classic Lexis (but not Westlaw), and I haven't checked whether those results differ.

More Details on Constructing Search Queries

I am no expert on constructing search queries—I have mostly learned through trial and error (see my point above about the need for explicit instructions on how to search). Here are some things I've learned—please let me know if anything is incorrect:

In WestlawNext, while you should start a search with "advanced:" if you want a Terms and Connectors search and aren't using an operator such as /p, this sometimes creates odd relevance-sorted results. For example, the first result for [ "doctrine of equivalents" ] is Warner-Jenkinson (a Supreme Court case on the doctrine of equivalents), but the first result for [ advanced: "doctrine of equivalents" ] is a district court case—Warner-Jenkinson is #22. But in other cases (when there was no common term like "of" in search phrases), it did not matter: [ advanced: "written description" & patent ] and [ written description patent ] had the same first fifty results, appropriately starting with Ariad v. Eli Lilly.

In LexisAdvance, although the "Search Tips" will tell you that a search for sources containing both "written description" and "patent" should be constructed as [ "written description" and patent ], the first result for this search is the Federal Circuit's Markman decision (not a WD case), and Ariad isn't listed until #132. Searches for [ "written description" patent ] or [ written description patent ] still start with Markman (the Supreme Court decision), but they have Ariad at #9 and #5, respectively, which is an improvement. Similarly, while [ patent & obvious ] starts with Graham at #1 and has KSR at #9 (two of the most significant Supreme Court nonobviousness cases, the cases in between aren't very helpful—including #2 Twombly and #3 Miranda (yes, that Miranda, which mentions "patent psychological ploys" and what is "obvious [about] such an interrogation environment"). The search [ patent obvious ] pushes KSR down to #22, but at least the cases in the middle involve patent law.

To search for cases on Google Scholar, click the "Legal documents" radio button. (After running a search, you can restrict to "Federal courts" if you want.) Results can differ when search terms are put in quotation marks (searching for an exact phrase) or not, although this did not have large effects on the results for searches I tried; e.g., Warner-Jenkinson is #1 for either [ "doctrine of equivalents" ] or [ doctrine of equivalents ].

While the order of search terms does not seem to matter on LexisAdvance, it sometimes (but not always) does on WestlawNext and Google Scholar. On Westlaw, the searches [ patent obvious ] and [ advanced: patent & obvious ] produce the same case results (with Graham at #1 and KSR at #4), but these are very different from searching for [ obvious patent ], which has Graham at #3 and KSR at #23. On Google Scholar, [ patent obvious ] has KSR at #1 and Graham at #2; [ obvious patent ] had Graham at #1 and KSR at #3.

Of course, the specific search terms chosen are also crucial. Searching for [ obviousness ] has Graham at #1 and KSR at #2 on Lexis and Google Scholar; and KSR at #7 and Graham at #14 on Westlaw. Searching for [ nonobviousness ] has Graham at #1 on all three but KSR at #27, #48, and #55 on Lexis, Westlaw, and Google, respectively (presumably because KSR only uses this term once, when quoting Graham).

3 comments:

  1. As a practitioner noted in an email, cost is also obviously an important factor for those who don't get free Westlaw and Lexis access. My post was just focused on who has the best relevance-sorting algorithm, but for practitioners who are choosing how to do legal research, the take home message might be that Google Scholar is not that different from the paid search tools for quickly locating some relevant cases—and it is free.

    ReplyDelete
  2. Surprised your results were so good with weslaw next. I never, ever get what I want listed first there - even if I type it in by name.

    ReplyDelete
  3. Interesting - do you have examples of queries that didn't work?

    ReplyDelete