Wednesday, October 17, 2007

Vivisimo hits the bullseye--Rebecca Thompson on Velocity 6.0

I'm attending a presentation at the Omni-Parker Hotel on Vivisimo's new enterprise search tool that promises to bring enterprise 2.0 to search. Rebecca is the VP of Marketing at Vivisimo. I attended a reception afterwords, but did not partake of the "Terminal Velocity" tequila drink.

Organon --"Where enterprise search is today."

Rebecca showed a sample search screen from pharmaceutical company Organon. As she rightly pointed out, most companies do not have a search this effective. Their enterprise search has content from a variety of domains (types of sources) including people bios / intranet / internent / sharepoint/ documentum / network. Like other Vivisimo GUIs I've seen, it has multiple sets of clustered results, one based on its semantic navigation, and two others by criteria (or metadata) specific to pharma, one by "therapeutic area" metadata and another by "brand name" metadata.

Next Steps--Enterprise Search 2.0

Web 2.0 is changing user expectations. People now want to interact with the information in a simplified way. If they have to take a training class, it's not going to happen. People go home to web 2.0. they use delicious, facebook, digg it, and share information (my informal surveys of our youngest associates suggest that these tools have not quite penetrated as far as Rebecca suggests). The new generation's method of communicating has been facebook--there is no need to send email when they can post. (This is certainly true--see my colleague Doug Cornelius' post on summer associates' use of web 2.0 technology). The sites are so easy to use that people won't suffer through training. Their common complaint is, "Why can't we have a corporate facebook."? They want "a better way to communicate and collaborate."

Enterprise search is a natural fit with collaborative software because the search goes across content domains and pulls in content from different sources.

Social search gives people power and control over their search, also leads to buy-in through their investment and participation in the process.

Three aspects of social search in Velocity 6.0

· Social Tagging
· Collaboration
· Social Networking

Social Tagging

There are three types of social tagging in Velocity 6.0 ("6.0").

The first and simplest option is "voting" on usefulness of results by clicking on a green "thumbs up," or a red "thumbs down." Votes can affect relevancy, or not, as you choose. There can also be some "super-raters" whose votes have more weight or who are the only people whose votes count. The system knows who you are, so you can change your vote (and can't vote multiple times on one document).

A second kind of tagging is "rating." One to five stars, and average rating is also displayed. Rating has the same set of controls on whether relevancy is affected. The engine can sort results by ratings.

The third way is adding keywords (I'd call this real "tagging"). The keywords become metadata bound to the search result and add context to the documents that you and others can see. The firm can control the available vocabulary for tagging--auto fill or drop-down list. Permissions to tag are controlled in the same fashion as voting and rating. Tags will not reveal documents otherwise marked as secure and documents are not revealed to a walled-off person just because they search on that tag. The engine will start to show similarly-spelled tags as you type. The 6.0 engine displays and finds tags by clusters on your search result.

Collaboration

Users can also annotate search results with their own thoughts and commentary. People's comments on documents remain as a type of institutional knowledge. This can turn into a view of people's work and experience.

Every user has a profile and can be alerted if there is a change in a search result or a given tag. This is a "reader-like" feature that can lead to more collaboration as users can see others that are working on similar issues or documents.

Virtual Folders

6.0 has "virtual folders" where search results can be stored. Folders can be public or private, although I did not see the "private" feature in the GUI of the demo. Individual search returns, that is, pieces of content, are dragged into that folder.

Virtual folders are shared on a group or public level.

One client of theirs, a "major NY-based media company", has a fixed group of KM / library researchers who set up folders on particular topics, and then grant project teams have read-only access.

Virtual folders appear only when relevant to your particular search.

One limitation, elicited by a questioner, is that you can't tag or annotate an external URL. Binding search results to documents rather than URLs allowed faceted guided navigation.

Social Networking

Search for personnel shows the tags that they've used. A search for content can also switch to a view of the people who have authored or tagged the content.

Clustering on employee results--after running a search for a particular key word, 6.0 can further cluster semantically on the employee information. This approach appears to be a great way to find particular types of employees with particular expertise.

Administration

As an administrative and KM tool, 6.0 has a set of tagging dashboards. They show top keyword tags, top taggers, by time frame. It helps manager see what employees are doing and demonstrate ROI.

The next step will be to give people more control over what they see in the results.

Other Thoughts

Rebecca recommended to treat search as an ongoing, living project. Don't set it up and walk away. Keep an eye on user behavior and on how content is changing.

Velocity 6.0 can be configured to ignore or stop clustering on certain terms, or to relate one "theme" to another, if, for instance, a company name was also an acronym.

Looking Back At What I Wanted

Last week I posted on the 6.0 announcement. There were a couple of features I was looking for:
  1. Pivoting on other users' tags and favorites;
  2. Pivoting around others' tags;
  3. Ranking or segregating people's tags on documents based on their role within an organization;
  4. Controlled sets of tags (as an option);
  5. Predicting and displaying others' tags based on what you have started typing;

The short answer is, the demo I saw absolutely had features 1, 2, 4, and 5. It also had the ability to rank the tagging based on a person's role in the organization, which is close to what I was looking at in feature 3 (you can have "super taggers").

Reaction

Vivisimo has really hit the Enterprise 2.0 target here. The four different ways that 6.0 brings in social collaboration into search have the potential to provide a rich, intuitive search experience inside the firewall, without much training required. Both tagging and virtual folders have tremendous potential to enhance collaboration and findability of content and people. And it's already associated with a really strong enterprise search engine.

The lawyerly skeptic in me wonders if lawyers will start contributing by rating, ranking or tagging, when a social collaboration tool like this is turned on. I think, however, that because a tool such as tagging has immediate personal benefit (i.e., it will help you find what you've previously found), without requiring significant investment of time, it will likely catch on. Tagging has even more benefit the more and more broadly it is used.

Taking advantage of the annotation feature will be more complicated and may require significantly more structure and institutional investment such as organization by practice areas, since lawyers by training can be quick to criticize and are proportionally more concerned about others' criticism of their comments.

Lynda Moulton on Social Search

It feels like there has been a tremendous crunch on my time since roughly mid-November. With the holidays approaching I am finally getting around to revving up some posts on this blog again, and clearing out some discussions that I have been saving up for a while.

I attended (in person) Lynda Moulton's talk introducing business motivators for social search , part of Vivisimo's Velocity 6.0 Boston event in Boston that I posted about earlier. Lynda's talk was quite dense but provided a "knowledge management" take on this collaborative technology; that is, she explained in terms familiar to knowledge practitioners the reasons why social search provides such an advantage over search without social context.

Lynda posted on the session as well.

Why does social search help?

Under the knowledge management concepts of "trust and validation," people will use most and seek to find content (such as a prior brief of a litigation expert) that has been validated by someone they trust. The extent of someone's trust in content is based on the content's expertise, authority, and affiliation as well the seeker's professional and personal relationships with the source or "voucher" of the content.

Seeking the company of others with similar business challenges is an existing, sound KM model for "bringing more to your work." Meeting in a trusting spirit and participating in sharing is known as a great way to get more out of content and your business.

How Does Social Search Help?

There are several different styles or methods of knowledge sharing.

  • Informing--reflected by commentary and analysis

  • Visualizing--done by dashboards, clustering

  • Demonstrating--reflected in clustering and federating

  • Expository--done by annotating and tagging

Together these tools play to our inclination to be self-sufficient with our technologies (and these tools work together in a social search enterprise tool). The human touch has much to offer clarity of knowledge. People are willing to engage in knowledge sharing when in fits in their workflow. You have to have early adopters, enthusaistic self-starters, and sharers. Once they demonstrate to others, it becomes contagious. Being able to see how others view a set of content, without having to ask them, is a new way of looking. We expect social search to play a strong role in organizing activities and work, at many levels.


              • for communities of practice--organize people
              • for domains--organize content repositories
              • for content--organize search targets, be they documents, video, web pages, or people's expertise
              • for clusters--organize content groupings
              • for tags--social tagging has no controlled vocabulary, metatagging is controlled.

              Can social search "transform behaviors into business wins"?

              Social search has the potential to elevate search discoveries into teaching moments:

              • it can leverage lost assets
              • it can save others time by placing discoveries in view with notations
              • nuggets of information can place content in context.
              (This is really important because context is the bane of the legal work product searcher's existence--you need to know what how the document's litigation / deal / transaction context differs from yours in evaluating whether to use a sample or form).

              Because of people's demand for trust, who was associated with a document can be just as important as what it is. (to provide an example, the brief on an evidentiary point authored by a partner noted for his learning on the subject is infinitely more valued than the one-off memo of the summer associate).

              Leveraging your network can save you time--annotating what you learn as you learn it increases your own understanding and that of others as well.

              Clustering has the effect of revealing relevant content no matter how you search for it.

              Tagging can prevent the "second fire drill."

              Benefits of these social behaviors accrue over time.

              Social Search Adapts Content To The Culture Of Organization

              Social search moves the content into an organization that members of a common culture understand.

              For instance, the pharmaceuticals industry is highly technical, with strong demand for efficiency, and its teams are made up of professionals with advanced degrees (I think Lynda suggested that taggers and annotations in this culture can afford to be quite technical and not intelligible to those without a Ph.D in biology.)

              In media, content is multi-media, poorly labeled. Teams are highly collaborative, and need in-depth and accurate fact-finding.

              Financial services pushes tens of thousands of products and funds out to its millions of customers. Customers have a wide variety of sophistication. Customers are also trading partners.

              Perhaps more collaboration and knowledge sharing in the defense industry would have led to a lower priced toilet seat.

              Recommendations For Social Search Projects


              Lynda recommended that a social search project needed to start with some questions about an organization's content:

              What are our most important knowledge assets?

              Could we benefit from collaboration?

              Why would content sharing get to better business processes?

              Lynda had several recommendations for implementation of any social search.

              • Build a map of who works and what content they use (I don't see how this is any kind of prerequisite for implementation of social search. Part of the idea is to let people build thier own maps of the content they use and see how they fit with others' maps).
              • Find teams with early adopter attitudes and with serious information gathering challenges;
              • Get a vision, a target that will give an edge;
              • Get a bunch of wins;
              • Communicate the outcomes and plan for the next; and,
              • Don't expect technology to solve the problem.
              You will need people who understand content and content architecture, especially on teams that might otherwise be skeptical.

              Metrics

              A woman in the audience asked if metrics for proving success are different with social search? Lynda suggested that the best metrics are stories. Search logs can also be a good source of success proof. You can set up a system where the expectation of how it is supposed to work is conveyed up front.

              ( I don't find this satisfactory--can't you look at amount of content, ranking, annotations contributed, to develop metrics and establish a return on investment? Another way to establish success would be to conduct pre-and post-rollout surveys of people's satisfaction with their ability to find stuff. The hazard there would be conducting the survey too soon since the value of this type of search would increase dramatically after people had been contributing content for a while).

              Friday, October 12, 2007

              Major enterprise search development--Vivisimo & social search

              Enterprise search vendor Vivisimo announced this week that they will be including a social collaborative component with the next version ("6.0") of their software. Lynda Moulton blogged about it; she is presenting next week at both a Vivisimo-sponsored webinar and also a more private event that I will be attending (disclosure: I may get a free lunch out of the deal). I have heard Lynda speak in the past and have generally been impressed with her depth of knowledge about search.

              Accessing Group Smarts Through Tagging and Ranking

              Providing some social context, through user tagging, shared user search result ranking, and exposure of other's favorites, has great potential to enhance findability and context inside the firewall, and I think 6.0 is a major development. I have previously posted an introduction to tagging, pivoting in Del.icio.us, and on CommonCraft's "howtoon" video introduction to social tagging.

              The leading example of leveraging mass opinion to enhance findability that I know of outside the firewall is Amazon. Amazon uses not just these three methods of social collaboration, but also now has designated key reviewers, user blogs, and much more, such as listing "what do customers ultimately buy after viewing this item?" But the key remains other people's reviews, which let you know if what you are considering buying is any good, and other people's tags, which help you find the products. It is no accident that other major evendors such as Tiger and even brick-and-mortar powerhouses like Sears have jumped on the customer review/ranking bandwagon. It simply works, for the users and the companies.

              I've been reading legal scholar Cass Sunstein's Infotopia: How Many Minds Produce Knowledge, which provides a clear explanation for the effect of "group wisdom" on decision making. He is quite hopeful, though not Pollyannish, on our ability to pool information. If people on average are more than 50% likely to rank and tag accurately (substitute "answer a question right" or "make a decision" as needed), then with a large group, ranking will tend to dramatically enhance accuracy, that is, the group will almost always identify "the right answer" or the most relevant page or document for the particular search. The outlier case is the converse; if users are not likely to be accurate (as when asked to guess the distance to the moon, or who will win the next Federal Circuit decision on obviousness leading to patent invalidity), then a large group will be even less likely than an individual to be accurate.

              The more people and the more content you throw at it, the better group intelligence will do. By contrast, interpersonal collaboration and intrafirm internal communication becomes much harder the larger (and more geographically scattered) the enterprise gets. Tagging, like other enterprise 2.0 tools, has a way of making the enterprise feel smaller.

              Other Social Software Inside The Firewall Vendors

              There was some fairly insightful coverage of Vivisimo's announcement in eWeek comparing Vivisimo's offering with a social collaboration module introduced by guided navigation pioneer Endeca. From a professional services perspective, I think it compares more closely with Connectbeam. Connectbeam provides social collaboration inside the firewall, organized around private, group, and company "topics," but does not itself have a search component; their publicity materials indicate, however, that they have connectors and have integrated with "major enterprise search engines" (by which they seem to mean Google and Fast.) Honeywell apparently implemented Connectbeam with Google Enterprise in March 2007.

              There was also coverage of 6.0 in Information Week.

              Another enterprise social software vendor, this one purely on the "tagging" side, is Cogenz. They have an excellent demo that serves as a good introduction to tagging (see also my previous post on Del.icio.us, a free web tagging service that can be integrated with Cogenz' product) I especially liked their promotion of tagging as the way to "tap into the collective intelligence of an organization by collecting, sharing and connecting around unstructured information."

              One question is whether these tools have a connector or can readily be integrated with Microsoft's Sharepoint 2007 intranet / DMS search. Many firms already have it, it's free, and some firms (like Sheppard Mullin) have used it as the basis for intranet searches.

              Features of 6.0?

              I'll be curious to see if, like Connectbeam, "6.0" will allow users to pivot on other users' tags and favorites, as well as pivot around others' tags. Will it be possible to rank or segregate people's tags on documents based on their role within an organization? A litigation partner might only want to look at sample complaints or settlement agreements that have been used or endorsed by another such person.

              I will also look for how 6.0 manages the tension between the organization wanting control over what the tags might be (and perhaps to drive them to certain content and away from other content) and the necessity to free up people to make up tags that make sense to them. I've noticed with my Delicious tags that it is all too easy to add a comma or use a slightly different word and thereby miss much of the context that tagging can provide. Cogenz handles this tension by predicting and displaying others' tags based on what you have started typing. I'm curious if more popular tags will be displayed first.

              Wednesday, October 10, 2007

              Someone Else's Live Blogging--Business Innovation Factory

              Through Erica Driver's interesting article on the limits and frustrations of starting up in Second Life, I learned about the ongoing Business Innovation Factory summit, October 10-11, which has as one component a set of intelligent blog entries at http://www.businessinnovationfactory.com/weblog/ .

              While a little far from the field of litigation, many of the posts are quite thought-provoking, such as Jason Fried's post on 37Signals, a software-as-service (SAS) company that allegedly, because it is not driven to do lots of fancy updates that customers will have to buy every year, can focus on keeping the software simple to use and effective. Hooray for that.

              I bet that there will be a lot more interesting reading coming out of BIF's summit.