Full article on Fulcrum vs. Verity:
erving Up Knowledge -- Products from
Fulcrum and Verity can help you tap
previously inaccessible information streams
By Jeff Jurvis
The rush is on. Knowledge-the corporate gold that runs
through the veins of every enterprise-is a hot commodity.
Corporate databases, E-mail stores, document collections,
workgroup repositories, and other information sources are
prime targets for knowledge mining-and companies like
Fulcrum Technologies Inc. and Verity Inc. are
repositioning their powerful information retrieval engines
to become the industry's leading knowledge servers. -
Clearly, knowledge is not easy to come by. Learning how
information, people, processes, and culture alchemize into
corporate knowledge is a huge effort, and
knowledge-server technology makes up just a small part of
the technical-knowledge infrastructure necessary to do it.
But knowledge server products from Fulcrum and Verity
provide the technical backing for a key first step-making
previously inaccessible or hidden information streams
open for exploitation. - Both companies are working hard
at acquiring or building knowledge processing and
retrieval components to extend their products even further.
Fulcrum and Verity will have to integrate their products
with knowledge generation and organizational
development tools and processes to fill out a robust
knowledge architecture. The products-Fulcrum Knowledge
Network 2.1 and Verity Search 97-take similar
approaches in presenting the tools that take a company
through the first few steps in building the knowledge
management infrastructure. Both are lines of products
available in different bundles at varying prices. Many of
the products are enhanced, repackaged versions of existing
products.
Layered Apps And Services
The search technology at the core of each company's
knowledge server is the same as that used in many
corporate intranets and document management
applications. Fulcrum and Verity have layered new
applications and services on top of and beneath the search
technology to create a framework for an enterprise
knowledge architecture:
- An information access layer that enables indexing of
heterogeneous information sources;
- A searching and indexing layer that provides many
different ways to formulate searches and analyze results;
- An application layer that includes packaged tools such as
agents and administrators, as well as open frameworks for
building custom applications.
Sources of information are accessed without any special
processing or conversions and without the need to import
the data to a special location. This helps make the
knowledge architecture noninvasive and easier to
implement.
As the result of a head-to-head comparison of product
features in InformationWeek Labs and the results of an
application case study in the trenches at a
financial-services company, I found Fulcrum Knowledge
Network to be a better solution under our conditions.
While the feature comparison actually seems to favor
Verity Search 97 by a nose, Fulcrum Knowledge Network
provided more of the flexibility and interoperability we
needed for our Windows NT 4.0-based Web application.
Verity, however, has better cross-platform reach.
We chose not to emphasize indexing and searching speed
and accuracy because both products have established
positions near the top of the performance charts and any
benchmarks we considered were arbitrary and not
meaningful outside the context of our case study.
All of the components of Fulcrum Knowledge Network
integrated well with our Windows NT development and
deployment environments-server components are
full-fledged, well-behaved NT services; application
components are solidly designed ActiveX objects and Java
classes; and Fulcrum leverages the NT security model for
access control that is seamless to the user.
Verity's design emphasizes cross-platform reach and
because of this the Search 97 products cannot take
advantage of any particular platform. For the most part,
Search 97 has a Unix feel to it-the application
programming interface is a list of C function calls, and
administrative functions are available only through a
clunky Web page interface.
The trade-off is clear-Search 97 servers run on many
platforms; Fulcrum Knowledge Network servers run on
Windows NT only (except for the ba-
sic search engine SearchServer, which runs on some Unix
variants)
Both Fulcrum Knowledge Network and Verity Search 97
share similar features, but often differ in implementation.
Each product's information access components can index
hundreds of document types, including all Office-type files
and just about anything else found in a typical corporate
environment. Fulcrum and Verity can also index E-mail
documents in Lotus Notes and Domino and Microsoft
Exchange mail stores. Verity outdoes that by interacting
with PC Docs and Documentum document-management
servers and Sybase and Informix relational databases.
Fulcrum calls its information access components
knowledge activators. These activators are intended to be
plug-and-play gateways to information sources such as
Windows NT and NetWare file systems, Web sites, Notes,
and Exchange.
Each activator bypasses the built-in search functionality in
Notes and Exchange and lets the indexer access the
documents directly. The Fulcrum knowledge server
maintains this index information in its own database.
Verity's server, on the other hand, will leverage an
information source's existing index. Search 97 can do this
because companies such as Lotus and PC Docs use the
Verity index format in their products. Because the
indexing work is done once and the results shared among
the products, there is no danger of separate indexes getting
out of sync.
On the other hand, Fulcrum users can choose to index
documents in whichever way they choose and can index
the specific information they want at the time they want.
Fulcrum has published its activator programming interface
and will rely on third parties to develop more activators
for products such as PC Docs and most relational
databases.
Searching And Indexing
The searching and indexing layers in Fulcrum Knowledge
Network and Verity Search 97 really show off the
state-of-the-art in information retrieval. The latest
advances include natural language parsing that
preprocesses a sentence into the usual combination of
words and operators that the search engine expects.
For example, instead of a user needing to construct a query
like "books AND Java AND teach" she could try a
sentence like "What books are available to teach me
Java?" If the results of a natural language query or a
regular query are not quite what she is looking for but
close, she could try Verity's "query by example" or
Fulcrum's "intuitive searching" to refine her search. Both
features take the results of one search and funnel them into
another search to try to fine-tune the query.
Another way to index documents and refine queries is by
document properties such as author, title, and revision
date. Accessing properties of our Microsoft Word
documents was easiest with Fulcrum's server. The Fulcrum
indexer automatically extracted and processed author, title,
and revision date. With Verity's server, I had to fiddle with
the wholly unintuitive style configuration files to get this
feature to work just the way we wanted.
Once a query is formulated and sent off for processing, the
knowledge server uses previously built indexes to match
document text and attributes to the query and returns a
result set. This is almost always a lightning-fast
process-only massive indexes require more than a few
seconds to search.
Results include a title or a file name and a link or some
other reference to the document itself. The document is not
stored in the index and is not retrieved from the file
system, the Web server, or any other document store until
requested.
Both the Verity and Fulcrum servers use summarization
algorithms that provide short overviews of document
content. These summaries usually consist of several of the
document's most significant sentences. This same
weighting scheme is used to assign relevance ranking to
each document returned in the result set. The ranking
techniques are contextual and not simply word hit counts.
Breaking The Mold
Knowledge server applications certainly do not all fit the
same mold. The most general kinds of applications are
those that provide a way to identify previously indexed
information sources and to construct search queries.
Verity and Fulcrum both ship client software that does
this.
Verity Search 97 Personal is a collection of
JavaScript-enabled Web pages that let a user connect with
and search any Search 97 server on the intranet, and to
index and search personal document collections on the
user's local file system. Search 97 Personal also includes
Verity's KeyView document viewer, which can display
more than 200 different kinds of documents without the
need for the original application.
The Fulcrum client software can be installed in all the
right places on a Windows NT or 95 desktop. You can set
it up to run as a standalone application-or, better yet, run it
in Microsoft Exchange/Outlook or Internet Explorer.
Fulcrum Find! works very well as a Microsoft Outlook
extension, and WebFind! is a nice example of how a Web
application can integrate HTML and ActiveX or Java
applets.
The rich WebFind! interface uses Fulcrum FulView to
open and view more than 150 different document types
with search-term highlighting to help you to quickly zero in
on the stuff you are looking for. WebFind! also integrates
Fulcrum's Knowledge Map, an Explorer-like tree view of
all indexed information sources, and provides a detailed
result list with summary information and document
properties. The best part is that WebFind!
is built from the Knowledge Builder Toolkit-Fulcrum's li-
brary of ActiveX objects, Java classes, and Web
scripts-and makes a great starting point for custom
knowledge retrieval applications.
Agent Services
Both Fulcrum and Verity have taken the logical step in
extending their search services into agent services. By
adding a user interface and a server database, their
knowledge servers can remember queries submitted by the
user, run them periodically, and return new results via
E-mail or a personalized Web page. Both products do this
well mainly because this kind of plain-and-simple agent
implementation just plain works.
But when you get beyond the plain-and-simple and need to
build a knowledge management application that fits your
organization and its unique knowledge architecture, you
need a well-designed, easy-to-use interface to the guts of
the knowledge server. Fulcrum did it right in Knowledge
Server version 2.1. The product's NT-specific
implementation exploits Microsoft's Component Ob-ject
Model (COM) and ActiveX to provide an
easy-to-understand interface to almost all knowledge
server functions. Visual Basic and Visual C++
programmers will have no trouble leveraging this library.
Web developers can use JavaScript, VBScript, CGI, Java,
and Active Server Pages to integrate knowledge server
components into Web applications.
Verity's developer toolkit is much more difficult to use.
Many knowledge server functions are accessible, but only
through a somewhat cryptic C API (which is fairly typical
as C APIs go). Web developers need to learn Verity's
SearchScript language to gain control over search-result
formatting.
The case study used for these evaluations highlighted
Fulcrum's openness and adaptability. The application I
looked at was Web-based with Microsoft's Internet
Information Server 3.0 running on Windows NT 4.0 and
Internet Explorer 3.0 on Windows 95 and NT 4.0.
My goal was to develop a tool to enable company
knowledge workers to explore corporate information
sources through a string of context-specific searches. For
example, when a user tries to determine whether or not the
company should enter a business agreement with another
party, the tool lets him build and run queries on specific
lines of business, types of agreements, and other criteria
typical to the company's business.
The knowledge server returns weighted results from many
different information sources including intranet and
Internet Web sites, Lotus Notes and Microsoft Exchange
E-mail stores, document- management systems, NetWare
and Windows NT file shares, and relational databases.
In addition to the open-ended query, the tool will also
deliver results proactively, based on the specific problem
under analysis. The company studied how the most
successful knowledge workers conduct their analyses and
what can be distilled into knowledge nuggets. The tool
uses these nuggets to augment the queries built by the less
experienced knowledge workers.
Conclusions
Fulcrum's COM-based object library was perfect for my
Microsoft-centered implementation. The flexibility of the
query and result set objects let me write Active Server
Pages that combine the results of Fulcrum queries on file
shares, E-mail stores, and Web sites with direct calls to
relational databases (through Microsoft's ActiveX data
objects). Search 97 cannot match this programming
versatility. But it does partly make up for that with its
ability to directly search on more information sources than
Fulcrum can and with Verity's Topic technology, which
lets you store predefined, preprocessed queries with the
knowledge server. |
