I saw an updated demonstration of the Interwoven Universal Search / Lexis Search Advantage (IUS/LSA?) search product in early December 2008. An excellent lunch by Radius was part of the inducement to visit.
IUS is powered by Vivisimo's enterprise-class search engine. I was pleased to be able to see a more fulsome demonstration of the IUS/LSA product following on the demonstration at ILTA, now with the addition of a potentially powerful social search component for the law firm market.
IUS/LSA is a feature-rich, complicated product. I don't mean to imply that it would be difficult for users to understand, although certainly getting them to use the social search features would be a challenge just like rolling out any new business process is a challenge. I've therefore identified some previous posts in the next paragraph, should you like some background information on enterprise search, social search, and work-product retrieval search.
I first blogged about Vivisimo's social search in October 2007, covering Vivisimo's initial announcement, Lyndo Moulton's framing of social search from a "traditional KM" perspective, and then a live demo of "Velocity 6.0" as that product is called. In August 2008 I also blogged about implementations of IUS (without a social search component) at three law firms; I also noted the rollout announcement for IUS / LSA, again, without a social component, largely by reference to my former colleague Doug Cornelius' highly favorable post on KM Space. For more see my posts with the "enterprise search" tag.
IUS/LSA combines "work product retrieval" and "enterprise" search. As quite nicely explained by Cindy L. Chick in this 2004 post on the subject, these two have been essentially different categories of search products, with the the former sharply focusing on a limited pool of internal (perhaps highly vetted) precedent, and the latter "federating" or broadly combing different pools or silos of enterprise information. Typical content "sources" for enterprise search include a document management system, intranet, and traditional enterprise database such as Expert/Aderant.
The rest of this post discusses particular features of IUS/LSA.
Case Linking and Validation
A key feature of IUS/LSA is its outstanding case validity check (through signals from Shepard's case updating service) and citing references links. The search identifies cases and statutes present in the content and takes that reference information to Lexis' enormous database, containing information how courts have treated each published decision (i.e., citing favorably, overuling, etc.). It returns a signal to the search engine, which is visible as a flag of some sort in an HTML view of the document from within the search (on a click).
Case validity tells attorneys reviewing work product that the cases (or statutes) are or are not valid; citing references provide an extremely efficient way to find or collect all of the work product on a particular topic, by linking all of the work product that cites to a particular case or statute.
The significance of case validation for busy attorneys cannot be overstated. Litigators are constantly weighing the strength of their arguments, based on the facts at their disposal and the extent to which those facts mesh with a legal theory. Legal theories themselves carry varying degrees of authority or persuasiveness—compare for instance the complete invalidation of slavery under the XIII Amendment to the US constitution with the Supreme Court rulings on the validity or invalidity of various race-based affirmative action plans. Case validation allows lawyers to tell at a glance if they might be treading on shaky ground in citing to a particular line of cases.
The "citing references" ability (my term, not theirs) allows researchers to quickly find other work product resources that cite to the same case law or statutory resources. Since lawyers making a legal argument ethically must cite to the leading authority in their jurisdiction on a given point, even if to distinguish its application in the present instance, finding other instances of case citation can quickly lead you to other instances where other lawyers have made the same arguments (or argued the same point from the other side!). Examining work product that cites to less significant authority can also be very productive, since this authority may be one of a line of cases on the your side of a particular issue.
Both of these features are found in West KM, although West KM is not currently bundled with an enterprise search tool as powerful, versatile, and social as IUS. In particular, the ability to cluster results on the case and statutory authority is extremely powerful.
The newest aspect of IUS/LSA is Vivisimo’s cutting-edge social search features taken from Velocity 6.0. These have been officially released, although I do not personally know of any law firms that have actually implemented these features in production (they have been available in the broader business community for something less than a year, I believe).
Once results are found, users can tag, rate, or comment. The main initial use of tagging is to help individual users refind their documents and create personal precedent collections. The magic of tagging is that each tag in turn enhances findability of documents (they directly effect relevance ranking) and expertise. Users can also tag into "shared folders," say by practice area.
Furthermore, users can identify who applied the tag and see all of a users' tags. Pivoting on known user information is a big advantage of tagging inside the enteprise (as compared to anonymous social tagging such as available on delicious).
One of the truly powerful aspects of integrating tagging into search (tags are applied at the time a search result is displayed) is that it integrates tagging into normal work flow. Users search all the time; they don't have to go a separate system to enter a tag, and the tags are immediately displayed and available for their own use the next time around.
I am a huge fan of social tagging . It is a key aspect of Enterprise 2.0, will greatly enhance findability for individual users, and then will only get better as more and more people add their own targeted markers of significance and relevancy. Ranking and commenting also have great potential for enhancing search inside the enterprise.
Like other enterprise-class search tools, IUS displays sets of results clustered by key metadata alongside a directly accessible relevancy-ranked set of search results. Where a collection contains significant metadata, drilling down into a set of search results through clusters of metadata is much more effective than a simple ranked list, because it lets users identify resources with directly relevant attributes or whose attributes are close enough to be useful. It also might let a searcher exclude large sets of returned documents.
For instance, a litigator looking for a Markman brief in patent litigation where guided navigation is available might search across all courts, but then drill down into the particular jurisdiction or into jurisdictions in the same federal circuit.
Semantic or Concept Clustering
In addition to clustering into categories predefined by DMS administrators, the Vivisimo engine has the remarkable ability to group documents into clusters based on concepts extracted from the documents. Vivisimo's "calling card" is its ability to generate clusters of documents based on similarities between groups of documents; at the same time, it identifies the groups by what it determines to be key characteristics of that "cluster." Depending on what the settings are, there can be clusters within clusters, allowing you to drill down into the set of documents most relevant to your search (instead of starting the search over when you get too many hits).
See for instance this clusty.com search for "senators", which lists "members" (of Congress) "Ottowa Senators" (the hockey team), "Legislation, Law" (which includes state senate information), and so forth.
The out-of-the-box people search looks at document authorship and official firm website biographies. Expertise can also be identified through looking at who is tagging what, which is exposed both through the lens of individual tags and through a view of a user profile. I suspect it would be also possible to use IUS/LSA to incorporate and analyze billing and matter records to further refine the people search.
Searching Across Silos
A key challenge for KM and information management strategy is the distribution of enterprise content across mutually exclusive “silos” of information. A matters database might have information about the general type of work being done on a particular deal and the partner who opened the matter; a document management system might have documents tagged with those same matter numbers; a separate system might contain final pleadings or vetted content; and a billing system might indicate who has worked on that matter the most. Typically the three sets of data never meet, and certainly cannot be leveraged in any reasonably user-friendly manner.
Enterprise search, of whatever platform, provides the “glue” that can bring all of these pieces together. A search for Delaware patent litigation briefs through such an engine might identify not just the Markman brief that the attorney was looking for, but also patent matters and people who with previous experience from that jurisdiction. In some use cases an enterprise search might even extent to designated or licensed content from the internet or proprietary databases.
IUS/LSA clusters based on source as well as documents' content and metadata.
IUS also has "Spotlights", very much like “Best Bets” or canned prepared sets of responses to particular queries. Knowledge management or other firm groups can promote particular sets of precedents or practice guide. In the demo, a search for "motion philadelphia" gets "How to file a motion in Philadelphia Court of Common Pleas."
This kind of spotlighting ties in well with efforts to provide procedural guidance for litigators, or deal checklist assistance to junior business lawyers. I am engaged on a wiki-based project to gather just this kind of information (on the litigation side).
The search previewed here will also add metadata about court and judge to that about cases and statutory references. This is through a trick known as “entity extraction”, where key parties or case numbers from the documents are matched up with party, case, and judge information from Lexis’ massive databases. I am not sure if this feature has been publicly released.
They also demonstrated automatic document profiling, through which documents can get assigned document types such as memo, pleading, correspondence, and so forth, based on their content and similarity to comparison sets of documents. I do not believe this feature has been publicly released either.
Some clients are using search to populate client pages that combine information from internal sources and saved searches of the web, in Sharepoint web parts. They have a federated search connector for Google, or, can search particular sites.
Such pages highlight enterprise search's ability to expose and make useful content previously buried in unconnected databases.