Finding IA at the Enterprise Search Summit

(this article originally appeared at iainstitute.org on June 20, 2011)

Last month in May, I had the pleasure of attending the Enterprise Search Summit East in New York City with IA Institute board member, Shari Thurow. Shari and I were on a quest to discover the role of information architecture in Enterprise Search. We didn't have to look too far, as both days were keynoted by IA Institute veterans: former IA Institute president and CEO of FatDux, Eric Reiss, on Day 1 and IA Institute founder and Principal and Senior Consultant at InfoCloud Solutions, Inc., Thomas Vander Wal, on Day 2 . Institute founder Bev Corwin was also in attendance and I quite was pleased to make a personal connection with a former coworker from PricewaterhouseCoopers, whom I hadn't seen in ten years.

In Reiss's keynote, "The Dumbing Down of Intelligent Search," he challenged search professionals to have the user, not the application, serve as the frame of reference for search. Using Google as an example, Eric showed how the algorithm may not provide the correct context. Those who build the algorithm need to ensure that contextual metadata is available in the CMS. Eric also challenged implementers to understand the business and educate the content providers of those needs. "Matching patterns is not the same as matching needs," he explained. And lest the users themselves forget their own power, Eric encourages all users to be critical and experiment, learn basic strategies and not to take for granted that the search solution is intelligent.

Thomas Vander Wal's keynote on Day 2, "The Search for Social," was a fitting bookend, showing how to deal with all the input once your Enterprise Search team has embraced the user. VanderWal described tools that go beyond searching for artifacts such as documents, emails and image/video content to searching for human resources, knowledge and expertise within the enterprise. Many presenters demonstrated social search tools for finding user profiles, activity streams and Yahoo! Answers-style knowledgebases.


A Common Theme

IA/UX was a prominent theme. Throughout the conference we noted terminology from the information architecture/user experience umbrella nestled within discussion of ECM, SEO, text mining, predictive analytics, policy and governance. Terms like information glut, findability, folksonomy, facets, and rich semantics, as well as a big focus on the user experience.

A major concern in the Enterprise Search community is the question of what exactly is new in search these days? Reiss noted that there has been no major new search engine since Google launched in 1998. Google Search Appliance and Microsoft Sharepoint are still dominant. According to a panel of experts moderated by Martin White, called "The Renaissance of Search," Enterprise Search has been running on autopilot for a long time and is only now finding innovation coming from places like mobile and social technologies. Panelist Alan Pelz-Sharp of Real Story Group said that consolidation around a product (Google, for example) does not equal maturity of a discipline. Panelist Hadley Reynolds of IDC, pointing to the now established mobile platform, said that the Google model is not ideal for mobile apps. For example, "A playlist model would work better for mobile search applications," he said. Innovative thinking around search for mobile should be a growth area.

As stated above, user experience was a huge theme at the conference. Panelist Lynda Moulton of The Gilbane Group highlighted improvement in user experience as a major new effort in Enterprise Search. She said that semantic technologies have been built on artificial intelligence platforms and wondered if it will "disappear like AI" or if they just need better UX packaging.

Focus on the user was refreshing but also pointed to a challenge. A theme I found running through many presentations was the sense that after 15 years, the Enterprise Search field is not marketing itself well as a discipline, both to business management and to the users themselves who benefit from search. Search managers feel they have to continually explain the value of search to users, which "…should be more common knowledge," according to Moulton. Enterprise Search is getting mired in a discussion of what the tools do, rather than what the search needs are and search managers are finding it hard to communicate the value of Enterprise Search, especially when search teams of 4 or 5 taxonomists (or often just one) have a small voice in the enterprise. Those in the information architecture world should find these challenges very familiar.

Exploring the Frontier

It doesn't take an information architect to understand that we live in an information jungle. Kathlina Phillips and Tom Lutmer of Wells Fargo touched on this in their presentation, "Taming the Wild West of Intranet Search." Kathlina and Tom described challenges from information glut, data quality/integrity, duplication, missing metadata, to security issues and the difficulty of unknown content sources. These problems sprung out of everything from mergers, outside contractors, multiple authors, lack of publishing standards and a large and growing number of data services. In a commonly expressed scenario, search is often an afterthought to content producers who may create a document assuming it is just for them or their team, not realizing its value to future users of a content system.

For the most part, the user is only passively creating tags. "SEO tends to refer to people who pay for results," says Phillips. Managers often don't care or realize that a document may need to be found later on. How do you get users to tag documents when they don't know it could be useful? Wells Fargo's solution is to tie results to how people do business through user query analysis. Facets can autocomplete via dictionary of terms pulled from actual queries. A careful content analysis allows the search team to add missing metadata, by pulling keywords directly from the document, using pattern matching to identify terms, cross-reference by date and author, and implementing override tools to correct miscoded data.

Gilbane's Moulton said "Technology should not get in the way of the intellectual process. If it does, it is not working." The solution she and many others suggested is an improvement in "user experience packaging."

It's a Jungle Out There

Over the two days of the conference, we noted a healthy debate within the search community on how to implement Enterprise Search, suggesting a developing maturity within the field. Between presentations on how to integrate Microsoft Sharepoint or Google Search Appliance, we found a number of organizations developing integrated solutions ranging from strict taxonomies created within these frameworks to innovative social mashups.

On the side of a stricter taxonomy was Sarah Berndt, a Taxonomist at NASA's Johnson Space Center. Sarah presented a case study of the agency's semantic system. Stating that "Government loves structure," she proceeded to describe the development of a taxonomy covering more than 50 years of space missions and dozens of separate teams and projects, each with its own set of artifacts and occasionally duplicate acronyms, labels and terms. Her method to sorting through this abundance of artifacts involves creating broad categories that map to an information architecture in a way that is easy to maintain and fits the users needs.

Sarah's procedure was in line with many other presenters: Study what the user does, interview subject matter experts including repository owners and content creators, and use the findings to build the taxonomy. Finally, use the controlled vocabulary to connect the "stovepipes" into an integrated view, allowing ease of use and effective search. While a proponent of involving the user in the development and maintenance of the search taxonomy, referring to her solution as a "monitored folksonomy," it becomes a relatively strict controlled vocabulary that users must adhere to in working with the system, at least until the next iteration.

Following the NASA presentation, was a strikingly different approach at Citigroup. David Palaitis of Citi Enterprise Search and Nathan Treloar of Ramp, a content optimization consultant, presented Citigroup's experiment with social tagging. David and Nathan described an enterprise environment consisting of 300,000 users, 70 applications and millions of documents. Looking for semantic approaches to enhancing and augmenting content, they asked, "Is there a simple grammar we can use to get people to tag items and concepts?" The simplest solution was found in today's social media platforms—hashtags, which were already employed in the various content delivery frameworks being used at the company.

The Citigroup team found that you can improve search results, as well as the content owner's desire to tag artifacts, by augmenting structured metatags with the freedom to add one's own hashtags. These hashtags can then be cross-referenced by metadata such as the author name and department to give semantic meaning to the artifact. The search development team can then use triples to tag concepts and replicate those tags to all documents associated with that concept, allowing those artifacts to appear in search results.

Gregory Greffenstette, Chief Science Officer of Exalead, in a separate presentation, also recommended using semantic structuring for optimal user experience (Is this an example of that? Are these two things the same? How are they related?) to get more informative output. Search applications are learning how to pull and organize structured information and map it into facets represented by graphs, tables and maps.

Out of the Zoo and Into the Wild

Bev Corwin, another IA Institute co-founder who joined me during these presentations, uses a zoology/naturalist analogy to describe the top-down versus grass roots camps.

The top-down methodology demonstrated by NASA employs the same level of monitoring and reporting as bottom-up methods, but defines a specific, central set of terms and relationships. In this methodology, the search manager acts as a zoologist who takes animals from the wild and puts them into cages. A top-down manager is doing analytics on term usage and use patterns, carefully integrating different domains into an overall ontology. The result is a fairly restricted set of terms and relationships that defines how an artifact is labeled and entered into the repository. Additions to the ontology are made during a periodic review of user behavior, similar to the way Webster's adds dictionary terms to each yearly edition.

The bottom-up methodology demonstrated by Citigroup is looser. It lives in the wild, and requires more maintenance to secure the environment, but it allows the user to define things in real time in terms that they use. The search manager acts as a naturalist, monitoring user behavior and analyzing outcomes without interfering with the user's ability to label and sort artifacts in his own terms. This method requires tools that allow the search manager to cross reference user-defined terms to a semantic vocabulary that is ever changing.

As Corwin said, information architects should aspire to be naturalists. She urged the IA Institute and its members to create a democratic model to integrate top down with bottom up in a way that maps out standards from all different frameworks, protects intellectual properties and privacy, and facilitates integration. "We have to preserve the wild. Just like protecting our rainforests, our jungles, our national parks," she says.

So You've Embraced the User, Now What?

The second day's keynote was presented by Thomas Vander Wal, best known for coining the term "folksonomy." Thomas spoke from the other end of the spectrum, where search managers embrace social search technology, and ultimately become mired in the challenges of this new frontier. So when you have embraced the users and their myriad acronyms, hashtags and other metadata, how do you integrate these pieces into an effective search system? And how do you get users to care?

Robert Boeri of Guident says that it is all about findability. Search managers need to continually evangelize and train the user. In theory this should be easier to do in an enterprise setting since the users are your employees, but it isn't always the case. Boeri noted that many presenters at the conference are advising people to include the users in the design and maintenance of search systems, but often the users themselves don't want to take part. They want IT to build the system and then expect it will work as well as or "just like Google." He suggests creating a metadata dictionary to help with confusion of tags and hashtags, as well as inviting unexpected allies like librarians, taxonomists, tech writers, business analysts and even attorneys (mitigating risk is a big issue) to help evangelize the importance of effective search.

As information architecture practitioners, we have the knowledge and expertise to collaborate with enterprise search teams to build interoperable search systems, through a careful study of user habits, content systems and governance. The need was clearly stated throughout the Enterprise Search Summit and represents a wonderful opportunity for engagement with enterprise teams. If you aren't already acquainted with your company’s enterprise search team, we encourage you to reach out and continue the discussion.