Doug Cornelius .com

Continuing my live-blogging from International Legal Technology Association’s Annual Conference. . . .

This was a sparsely attended session (it was during lunch time) from Chris RuBert, Director Customer Care of Interwoven giving a Worksite 8.3 Field Update and Shawn on preparing for Worksite 8.5.

Technote 62008-2619 provides planning guidelines. For every 5 million documents in Worksite, you need a 4 CPU/Core server with 8-16 GB RAM. Within the guidelines, you want to be able to process 1 search per CPU/Core per second with an average search response times of 3 to 5 seconds. Investing in hardware will yield a return. More processors get you faster indexing and more searches per second. More RAM gets you faster searches. 64 bit is the recommended platform, with better performance and greater stability.  The Indexer should not be on a virtual machine. The concern is the IO bottleneck on VM platforms when under load. It is a resource intensive application.

The 8.3 upgrade is a little easier than past upgrades.  Get your hardware and point it at the production data.  You can identify potential content issues in the test and test real user searches against a live index since you are setting up a parallel system. You hit the switch to move from verity to velocity.

Indexing performance tuning guide is available on the Interwoven Support Site (72008-2659). You do need to get the virus and backup software out of the way of the Indexer.

Wildcard searching has changed. A custom dictionary is recommended. Wildcard searching requires 3 characters. Searches with dashes is a problem, but can be configured in syntax. Leading wildcards should only be used when needed. (We have noticed this to a be a problem. This is a big change in functionality.) The old setting automatically inserted leading and ending wildcards. With the velocity search engine, you want to disable the leading wildcard. The trailing wildcard is addressed with stemming. This is why you need to build the dictionary.

The current problem is that the indexer either indexes all versions or just the latest version. If you index all versions, then all versions come back in the search results (This is terrible see: Interwoven Express Search and Worksite 8.3 Update – Back to Wow!! ) The service pack will just show the lateste version that meets the search criteria.

Version 8.5 is trying to enhance the offline experience, email management, searching, MCC and multi-language support.

They are looking to have seamless online/offline switching. They want to handle a network failure allowing wired to wireless switch. They also want to have better synchronization and have it in the background. You can pause, see the progress and manually stop it (if you have a bad connection). Of course they want better performance. They are using a new SQL express client database and event driven architecture.

When you deploy Offsite, there may be a big initial synchronization which will put a big load on the server. You want to phase the rollout.

Offsite is going to be tied into email management.  You can continue filing email when you are not on the network.

Version 8.5 will allow multi-byte languages (like the Asian languages) in description, comments, to, from, cc and bcc.  Version 9 will be fully unicode.

Version 8.5 will also index workspaces, including the metadata and descriptions. It will enforce filing locations so that documents must be in a folder in a workspace. It will introduce a search for workspace in the save document window, giving the workspace in the window. It also will not show the contents of the folder anymore. Showing the documents takes time to render. There is a button to show the documents. Seeing the documents helps you to see if you are saving the document in the correct place. Since version 8.5 will be indexing workspace, you need to make sure you have the recommended hardware.

Version 8.5 will allow server-side email filing. This will result in a big time saving. Currently any email filing takes at least a second per email. This is because the filing is happening on the machine. (It is fast on outlook/exchange because it is happening on the server.) Now the email movement will be a server-based transaction. The heavy lifting happens on the Interwoven Communication Server. They will using the email icon in Outlook to indicate that the email has been sent for filing and then again when it is actually filed.

There is also a new filing toolbar. When you highlight the email, it will offer a list of suggested workspaces. (Just like the Blackberry file email suggestion.) There is also an option to (i) file, (ii) file and delete from Outlook and (iii) file, print and delete from Outlook. The toolbar has an button to save the attachment into Worksite.

There is also an Outlook folder sync function. You can set up a link from your email folder to a Worksite folder. The user can focus on the organization and not have to focus on finding and filing. It will also work from any email access: Outlook Web access or Blackberry. A benefit of using a convenience copy in Outlook/Exchange is you can still use flags and follow up indicators.

You can also do delegate filing. If the secretary is filing the email, author will be the attorney not the secretary.

One issue is the physical location of the Exchange servers and DM servers. Since you do have server filing, you can put a load on the WAN.

The upgrade from 8.3 to 8.5 will require a re-indexing of the document collection.

Interwoven announced the winners of their MCC design contest at International Legal Technology Association’s Annual Conference.

Congratulations to Lisa Gianakos of Reed Smith LLP for getting First Place!

Continuing my live blogging from International Legal Technology Association’s Annual Conference, I sat in the session from Interwoven and MicroStrategies on using their Matter Centric approach to their document management system.

MicroStrategies and Todd from Lownestein Sandler presented their view on how Lownestein Sandler use Matter Centricity.

Specialized practice groups operate differently. Managing partners and other heavy hitters who operate on their own. Administrative departments also have a different way of dealing with documents. They do not work on matters.

You need to dealt with complaints that matter workspaces are organized by matter numbers. Attorney do not remember attorney numbers. They remember client and matter names. Lownestein had half of the firm wanted it organized by client name and the other half wanted client number.

They used shortcuts to see matters in a different way. Lowenstein organized matter workspaces named Client#, Matter#, Client name and matter name (in that order). They used a shortcut with a new name Client name, matter name, client#, matter#. They use a tool from Microstrategies that automatically creates the shortcuts.

The firm had a client with hundreds of matters with several opening per day. They really needed a way to organize the matter workspaces. It also caused a huge screen real estate problem. The solution was to build out a tree structure of the matter workspaces for that client using shortcuts for a fictional user account. The tool automatically adds newly opened workspaces into the client’s shortcut tree.

One user wanted to just fill out the document profile and not have to find the matter workspace and folder. They created a script that finds the documents and moves them into document folders. It also required a setting on the computer to keep the classic view and not the matter workspace view when saving documents and finding documents.

Then next problem is cross-disciplinary matters, where you multiple practices involved. For example, when you have tax, labor or another group working on a securities matter. They created an email address that will create a new folder, with the name of the folder in the subject line. You use the “file and send” function and file the email in the matter workspace. This allows the user to create a new folder without the elevated rights that would allow them to edit the workspace.

My ILTA Schedule

At the ILTA conference, Lexis and Interwoven announced that they have teamed together to provide some integration. (This is the big announcement I mentioned earlier:  Interwoven – Big Announcement at ILTA.)

Lexis Search Advantage is a new product from Lexis Nexis. It links to case law, statutes and regulations, with real time Shepard’s indicators for cited cases.  It also works with transaction documents by creating a virtual table of contents, a search for reusable clauses and links to companies and people.

The key to Lexis Search Advantage is that in integrates into Interwoven Universal Search.  Lexis is adding value to firm’s existing content and not creating a separate silo of information.  This collapses searches of internal content, Lexis content and updating cases into one platform and one step.  Internal content is combined with Lexis sources.

According to Doug Stansfield of Lexis, they partnered with Interwoven Universal Search because they thought it was the best enterprise search product in the marketplace for law firms. 

In the search results, there is a “research preview” button. This brings up a HTML preview of the document with live links to the cases cited in the document and Shepard’s signals to the treatment of those cases. Clicking the Lexis content triggers the Lexis charges. The charges are based on your firm’s subscription model. You also have highlighting that shows the search terms in the document. The Shepard’s signals are updated when the document is opened and rendered. Not when the document is indexed.

Lexis also allows a search of web content to be another place that you search with Universal Search.  I noticed the search they used also included content on JD Supra and some blogs. (Yet another reason lawyers should be blogging and posting documents to JD Supra.

For transactional documents, the document has a link to information on people and companies. If the document has IBM as a party, you have a live link to the Lexis Dossier on IBM.  You can also pull up the SEC filings for public companies. The links are configurable, so you could link to Interaction information or other internal sources of information instead.

Search Advantage also offers some auto-profiling of documents.  It can be matched to a firm’s taxonomy or a Lexis taxonomy.

There are three big advantages to the mashup:

• Improves ‘Findability’ of content in Interwoven repositories by enriching documents with external meta-data at index time. During indexing, Search Advantage scans the contents of all documents and e-mail.  When it finds a citation to legal authority (court case, statute, etc.), it can add the court or the judge to the searchable metadata of the document. This allows you to find, for example, all documents that cite a particular case or that were filed in a particular court.
• It eliminates several manual steps previously required to assess current legal validity of precedents and previous firm documents via ‘real-time’ Shepard’s signals. When a user previews a document the document will be shepardized in real time.
• Makes it easier to navigate from a internal document to case-law on lexis.com. Citations to previous cases, statutes, etc. are automatically hyperlinked to the corresponding source document on lexis.com when previewed from Interwoven Universal Search.

Universal Search is already an impressive product. This integration makes it that much more powerful.

There is public demo of the product on Wednesday in Texas Ballroom 5.

Press Release: Interwoven and LexisNexis Team Up to Introduce Integrated Solution

Knowledge workers spend approximately a quarter of their time searching for information, but how successful are they at locating what they are looking for? Our panel members have had enterprise search engines implemented at their respective firms for over a year and discuss the changes they have encountered with enterprise search.

Speakers

• Robert Guilbert – Wachtell, Lipton, Rosen + Katz
• Jeff Rovner – O’Melveny + Meyers LLP
• Rachelle DeGregory – Sheppard, Mullin, Richter + Hampton LLP
• Chad Ergun – White + Case LLP

My Notes:


Recommind at O’Melveny + Meyers LLP

Jeff was heavily influenced by the Long Tail (as explained in the The Long Tail by Chris Anderson). Their analysis of enterprise search tools was based, in part, by what was coming out of the consumer internet. The firm chose Recommind. Their search solution imputes lots of of information about a document based on the client/matter designation assigned to the documents. They pull information form the financial system, the matter tracking system, etc. and add it to the metadata for the document.

They started the Recommind proof of concept in November 2006, finished this June 2007 and launched it in September of 2007. They started with Recommind as a stand alone application. They used flickr as model for the the visual landing page. They also modeled the search training on searching for products on internet shopping sites. If you could shop online at Macy’s, you could use Recommind.

Their second stage of Recommind was integrating it into the intranet. For example, the people search uses the Recommind people search tool. You can filter the search results or used an advanced search to find very specific skills.

Autonomy for White + Case

Chad’s experience was similar to Jeff’s experience. they put together a very long list of features and comparison of four vendors. They picked autonomy. A big issue for them is that they have 38 office with many different language.

They did a quiet roll out of the product. Their IT systems are very decentralized. Each office had their own document managment system. It would take hours for an attorney to hook into all of the different offices and conduct the search across all 38 systems.

They have Autonomy index each of the systems and create a united search. They get blazing speed. (Especially compared to the searching each of the separate document management systems.)

Autonomy also has a desktop application to go along with the web-based search. This was really fast. It also can be incorporated into MS Word. As you type in a document it show you other relevant information in the firm’s resources.

They will also have a search for voicemail.

. . . .

Tagging

The panel thought the tagging features in the next versions of Universal Search and Recommind will be very useful.


Other thoughts:

• You may want to check out the AIIM‘s Market IQ Report on Findability: The Art and Science of Making Content Easy to Find

My ILTA Schedule

UPDATE: slight revision to protect the innocent

I just got a preview from Interwoven of a new product they are getting ready to launch. Interwoven is announcing the product at the International Legal Technology Association’s Annual Conference on Wednesday, August 27 at 1pm in the Interwoven demo room (Texas Ballroom 5).

It should be a great tool, adding a great deal of functionality to law firm users of Interwoven’s Universal Search. I will share more after I am authorized to go public.

My ILTA Schedule

In my last post (Interwoven Express Search and Worksite 8.3 Update) on Interwoven’s new tools I noted quite a few problems. Now, after re-configuring some settings we are getting back on course. Problems 3, 4, 5, and 6 have all been fixed and the searching is great. Those quirky search errors are gone. Express Search even seems to be running faster.

Express Search introduces a Google-like search for our document management system. It moves the search dialog from a screen full of database fields to one simple box. It also brings relevancy to the search results.

Going back to my earlier theory of searching, there are four basic types of searches for documents: fetch, recall, research and precedent. The Interwoven document management system has always done well with the fetch and recall types. Those are the types of searches that you know what you are looking and are the type of searches a document management system was built for. But it had always been fairly poor at the research type where you do not know what you are looking for. The search results do not come back ranked in relevancy so there is limited ability to deal with a long list of results. Also, even though we separated out document collection into several different libraries to speed the search results, the libraries have just continued to grow. The big libraries make the full text searches very slow in all versions prior to 8.3. Express Search brings speed and relevancy ranking. That makes it a great solution for doing a research type search. The sister tool to Express Search is Data Miner. It allows you to group the search results based on the metadata from the document profile. This gets closer to addressing the needs of a precedent search.

We are still testing and banging on the tools, but two problems remain on our list.

First, 8.3 is still configured to return all versions of a document that meet the search criteria and not just the latest version of that document. For key transactional documents, we will go through several versions of the document. We have our 8.2 document management system currently configured to only show the document a single time in the search results. That is problem 1 from my prior post.
This multiple versions problem is still a big problem. Unfortunately, it carries over to Express Search, DataMiner and DeskSite.

Second, it is a challenge to search by the document identification number. If you put the document ID in the Express Search box, without specifyiing the “doc.num:” syntax, it will execute a full text search and you may get the document back in the search results if you have included the document ID in the text of the document. (We generally put it in the footer of the document, but not always.) Even though you can search by document number there is no way to limit the search to a particular library and the search results do not identify the library of the document. The complicating factor is that we have six libraries in the document management and do not make the document ID unique except within a single library. So document number 123456 would be unique within a library but could exist in each of the six libraries. Of course, you could use the bulkier DataMiner which does display the library.

We can probably cope with document ID problem by keeping the old DeskSite interface on the desktop. People would use that interface to search by document ID.

But the multiple versions problem is still an impediment to our moving forward with Express Search and Worksite 8.3. We find it very jarring to the search experience.

I heard rumors that Interwoven is working on a patch to fix the multiple version problem and the library problem for the document ID in Express. In the meantime we will continue testing.

The document management system is the assembly line of large law firms. We sell our knowledge to our clients. But we package and deliver a lot of that knowledge in documents. I keep a close eye on the document management system to make sure it is keeping pace and hoping that it will show some functional improvements in this 2.0 world.

Our document management system vendor, Interwoven, has been throwing out some good ideas on how they are going to improve their product. Back in February, we ran a quick test of Interwoven’s new Express Search tool: Interwoven Express Search = WOW!! Last month I presented a Case Study on Worksite 8.3.

After a lot of load testing, checking configurations and hardware upgrade, we sat down to test them again.

The concept is still great. They stripped the cranky old search engine and plugged in a fancy new search engine from Vivisimo. They also combine several of the metadata fields into one search box, so you get a much more Google-like search. The search is still very fast.

In our current tests we ran into a lot of issues. We are still trying to figure out if is our problem or Interwoven’s problem or some combination of both. These were the problems:

1. Multiple Versions. If there are multiple versions of the documents (which almost always the case for our important documents) each version comes back in the search result as a separate item. Our prior configuration was just a single entry for the multiple versions.

2. Document ID. One of the core functions of a document management system is the fetch search. You enter a single unique document identifier and that one document is returned. We operate with multiple document libraries, so our document identifier numbers are only unique in each library. The identifier will be in the system multiple times, but can be limited by library. With DeskSite product from Interwoven, the search results show the document library so you can easily distinguish the document you are fetching for. In Express Search, the search results merely show the document application symbol and document name. That means you can’t distinguish the document from the list presented. This is especially problematic when you combine this the multiple version problem. Today we had a new problem. When we search by placing the document ID in the Express Search window, the search just fails to return results.
3. Exact Phrase. Putting quotes around a phrase is a powerful search technique. For some reason two word phrases worked fine, but three or more words in the phrase failed. “one two” works but “one two three” does not.
4. Three Letter Words. The search engine seems to ignore words with three letters or less. When the three (or two) letter word is put in quotes to search for a phrase, the search fails.
5. Wild cards with numbers. We configured the application to automatically put wildcards around searches in DeskSite. 1995 proxy fails.
6. Differences Between Data Miner and Express Search. A search for “1995 Proxy” worked in Express Search, but not in Data Miner.

Some of these problems may be easy to fix configuration issues. Or they may be a systematic problem with the new tools. Either way, we are probably pushing back the deployment date for Worksite 8.3. That also means we are pushing back the deployment date for the Matter Centricity tools as well.

Rizwan Khan talked about change management, in particular in moving to matter centricity. Moving from DocsOpen to iManage was a change in the technology not a change in process. Matter Centricity is a dramatic change in process and user behavior. Attorneys are willing to share files but not as willing to share structure. People want to organize documents in a way that makes sense to them.

Simplicity is key. The change is going to be severe, so make it easier for the attorneys and legal staff. Only push out a few folders at first. One firm said that the have selection process at matter opening to pick the folder structure. Another firm said that they allowed the attorneys to create their own folder structure.

Communication and training is key. Communication up front is key to clarify the business need and convey those needs to the firm-wide audience. You need multiple trainings. People will remember some features and not others. They will have questions.

Rizwan showed an eLearning tool for the Interwoven products. Unfortunately, I could not find it on their website. (Hmm! Why not?)

Rizwan pointed out how the Universal Search can used to deal with centralization, or lack there of. The Universal Search can pull from the disparate systems. Centralization benefits the IT staff, but provides little benefit to the attorneys.

(Universal Search would be very useful tool for mergers. Just point the search at the two firms DMS systems. You can work on merging the systems later.)

One plan for Universal Search is to also incorporate external repositories.

The indexing and search results is much faster in 8.3. A repository of 9 million documents took 30 days in 8.2 and 5 days in 8.3. The search results are much faster in 8.3 (I found this to be very true in our testing.) There were some charts that the user experience for tolerating delays is about 5 seconds. Anything longer than that rapidly becomes unacceptable.

I was asked to speak about The Firm’s Worksite 8.3 implementation. Kevin Hicks from Interwoven was able to supply the technical side of information. My main focus of presentation was on the user interface and user experience.

Everyone asks us for Google. “Why can’t we just have Google search our documents?” Larry and Sergei are billionaires because they figured out the special sauce for searching and ranking webpages. What our people are asking for is a simple interface that returns documents in a meaningful way. Instead, with Interwoven we give a lot of fields to search in very database look with documents returned in a flat result list.

Before throwing Interwoven under the bus, there are four types of searches and Interwoven does two of them very well. The four types are fetch, recall, research and precedent. Interwoven excels at the fetch and recall search. These are by far the most common searches run at most firms.

It is the research and precedent that deliver the knowledge management value by allowing attorneys to find and reuse relevant content.

Until Worksite 8.3 and Express Search, they did poorly at research. Interwoven Express Search now delivers on research. Nobody does a precedent search. That is one of goals for enterprise search.

A deeper discussion of the four types of search.

With the fetch, you have exact identifying information. For instance, with a document in the document management system you have the document number, or you have a filename and path, or a URL. This is core document management activity.

With the recall search, you have some distinct information about the nature of the item. You remember a matter it was associated with, who created it, when it was created, etc. With this type of search you typically get back several or many items and you need to sort through the results to find the item you were looking for.

The research is the type of search that an enterprise search was built for. You want to find information on a topic and you may have no idea if the enterprise has any information on that topic. Information could be stored in a variety of sources/databases.

With a precedent, the information that makes the item relevant is generally not in the text of the document. For instance, if I were looking for a purchase and sale agreement for a retail shopping center in Florida that is buyer favorable. The words “Florida” “retail shopping center” and “buyer favorable” may not appear in the document and if they do they may only appear once or twice. To enable this kind of search you need to harness the document collection to another database of information.