Enterprise search is a hot topic for large law firms. 

I recently cited John Alber’s excellent article on enterprise search. Another leading thinker about search is Oz Benamram, the mover behind Morrison & Foerster’s AnswerBase (AnswerBase article; AnswerBase demo).

In a message circulated to an ILTA listserv, Oz wrote about two additional search considerations (quoted with permission):

1) Search Engine vs. Enterprise Search:
John focused on the differences among search engines. Even even the most accurate search engine, however, is likely to yield too many results merely because most firms have so many responsive documents. What good enterprise search systems, such as Recommind’s, do to solve that problem, is to provide users with context to (i) further narrow down the results list and (ii) determine the value of each result.

For example, after searching for a “motion to dismiss” you may want to narrow down the results by jurisdiction, court, or judge, and select only those that won. You may want to focus on motions written by a specific individual or by the author’s practice group, or that were used when Bryan Cave was on the other side.

Only an enterprise search system (as opposed to merely a search engine) can connect all the parts of the information within the organization to provide you with the context you need. That’s what makes information actionable.

2) The Relevancy of Relevancy:
It is worth noting that relevancy is subjective. It depends on the user who runs the search, and the task at hand. For a litigation partner in the New York office, a document written by a colleague in New York is likely to be more relevant than a similar document written by a tax paralegal in Tokyo. Similarly, the documents most relevant to a lawyer writing an appeal will likely not be the same as those most relevant to a lawyer writing a merger agreement.

Few search vendors are addressing these aspects.

To accomplish what Oz refers to in point 1, the search tool requires “entity extraction,” that is code that allows it to determine names, courts, jurisdictions, dates, parties, and other relevant information from within a document and then present those choices as “search facets.” A facet let’s you select a particular attribute, for example, jurisdictions or deal types. For point 2, the search technology needs to give extra weight in relevancy ranking to the user’s own works and perhaps, the user’s own searches, or other indicia of the user’s interest.

Search has come a long way from “inverted indexes,” Boolean expressions, and proximity!