A Large-Scale Hypermedia Document Management and

Document Sample
A Large-Scale Hypermedia Document Management and Powered By Docstoc
					                               A Large-Scale Hypermedia Application using
                              Document Management and Web Technologies

                                             V, Balasubrantanian                             AlfBashian                      Daniel Percher
                                               President                                    System Architect                System Architect
                                            E-Papyrus, Inc.,                                  Merrill Lynch                   Merrill Lynch
                                          63E Reading Road                               400 College Road East           400 College Road East
                                       Edison, NJ 08817, USA                           Princeton, NJ 08540, USA        Princeton, NJ 08540, USA
                                       Tel/Fax: 1-908-548-7868                            Tel: 1-609-282-4717             Tel: 1-609-282-4722
                                         bala@e-papyrus.com                               alf_bashian@ml.com            daniel_porcher@ rnl.com



ABSTRACT                                                                                          Within Merrill Lynch, the Private Client Group provides
Merrill Lynch has initiated a major effort called the Trusted                                     planning-baaed financial management services to over four
Global Advisor to provide instantaneous access to current                                         million households and 225,000 small to midsize businesses
financial information to about 20,000 financial consultants                                       and regional financial institutions. Serving these clients is a
and other professionals across the corporation.       As part of                                  network of over 13,700 financial consultants and private
this effort, marketing information         about products and                                     bankers located in more than 550 sales oftlces worldwide.
services will be delivered to financial consultants, clients,
and the general public through an intranet and the Internet.                                      As part of a major effort called the Trusted Global Advisor
A number of researchers have reported on the requirements                                         (TGA),     Merrill    Lynch   is focusing      on providing
for industrial strength hypermedia. In this paper, we present                                     instantaneous access to current financial information to
a case study on how we have designed a large-scale                                                financial consultants and other professionals.     To support
hypermedia       authoring     and publishing    system    using                                  this effort technologically, Merrill Lynch is replacing its
document management and Web technologies to satisfy our                                           workstations used by financial consultants.      Workstations
authoring, management, and delivery needs.          We describe                                   connected to a mainframe with a character-based front end
our systematic design and implementation            approach to                                   are being replaced with a client-server        system and a
satis~    requirements      such as a distributed      authoring                                  Graphical User Interface (GUI) Shell. The design rationale
environment for non-technical authors, templates, consistent                                      behind the TGA Shell has been described in [10]. TGA has
user interface, reduced maintenance, access control, version                                      provided Products and Services Marketing (PSM) groups
control, concurrency control, document management, link                                           with an opportunity to re-design their on-line marketing
management,       work flow, editorial     and legal reviews,                                     system and leverage the full range of marketing materials
assembly of different views for different target audiences,                                       available in a variety of media.       HTML and Portable
and full-text and attribute- based information retrieval. We                                      Document Format (PDF) documents will be delivered to
also report on design tradeoffs due to limitations with                                           financial consultants’ workstations via an intranet. Subsets
current technologies.      It is our conclusion that large scale                                  of this marketing information will also be delivered to
Web development should be carried out only through careful                                        clients via a dedicated online service and to the general
planning and a systematic design methodology.                                                     public via a public Web site. As part of this initiative, the
                                                                                                  systems development team was assigned the task of building
KEYWORDS: WWW, Document Management, Workflow,                                                     a document repository with facilities for a group of
Systematic Hypermedia     Design, Distributed    Authoring,                                       distributed authors to create content, submit it for editorial
Publishing, Views, Templates, Information Retrieval                                               and legal reviews, assemble and publish information
                                                                                                  simultaneously and periodically to the three aforementioned
1 INTRODUCTION                                                                                    audiences.
Merrill Lynch is a financial-management    and advisory
company with a global presence in more than 40 countries.                                         In this paper, we present a case study on how we have
Permission to make digitablrard      copies of all or   part of   this material for
                                                                                                  designed a large-scale hypermedia authoring and publishing
personal or classroom use is granted without fee provided that the copms                          system using document management and Web technologies
m-e not made or disbibuted      for profit or commercial     advantage,    the copy-
right notice, the titfe of the publication    and its date appear, and notice is
                                                                                                  to satisfy our authoring, management, and delivery needs.
given that copyright     is by permission    of the ACM,   Inc. To copy otherwise,                The remainder of Section 1 describes the existing system
to republish, to post on servers or to redistribute      to lists, requires specific
                                                                                                  which delivers products and services marketing information,
permission    and/Or   fee.
HypmtexI                UK
             97, Southampton                                                                      its limitations, and the requirements for a new system. In
O 1997 ACM 0-89791 -866-5 .,..S3.50
Section 2, we provide an overview and details of our             recreate lost material.    Also, there is no way to ensure that
systematic design methodology.      In Section 3, we present a   the material is current.
review of related work in the areas of systematic hypermedia
design, document      management,     workflow, and WWW          The material       in the on-line     marketing      system    is
issues. Section 4 describes some of the design tradeoffs and     supplemented by an average of nine-pounds of printed
limitations of current technologies. In Section 5, we present    material delivered to financial consultants each week. This
our contributions and conclusions.                               includes brochures, marketing guides and newsletters. Much
                                                                 of this material is never used by the financial consultants.
1.1 Existing System and Limitations                              Any     multimedia    marketing    materials     the   financial
The existing on-line marketing system is a text-based system     consultants wished to present to their client, such as CD-
developed in a CICS/COBOL           mainframe environment.       ROMS or videotapes,        are obtained      through separate
Each financial product has its own chapter consisting of         channels.
about 24 documents on individual topics related to the
product, There is little consistency in the organization of      1.2 Requirements for New System
product chapters, with similar topics often having different     The business community not only wanted        to correct the
names in different product areas (e.g., Bulletin Board, Hot      deficiencies of the old system, but also wanted to facilitate
News, Announcements).      Many documents are a series of        delivery    of multimedia     material    in an integrated
about 99 screens of text consisting of 14 lines of 58            environment.       They presented      the following     new
characters displayed on a 3270 emulator. The text in the         requirements to the systems development team:
40,000-plus screens ‘me all displayed in upper case which is
hard to read.                                                    q   handle large amounts of unstructured material in
                                                                     multiple media
Documents are organized into a rigid but shallow hierarchy       q   provide an information structure that is consistent and
which, in a number of instances, require authors to                  predictable across all product areas
inappropriate y place subordinate material at levels equal to    q   develop an intuitive user interface and navigational
the principal material. Because of the system’s inability to         mechanisms throughout the system
provide linkages between documents,          authors had to      q   provide links between related materials
duplicate information. The hierarchy could be traversed by       q   present information in an easily digestible format using
a menu system.       The items on the main menu consists             graphs, charts, bullets etc.
primarily of the main categories of financial instruments.       q   provide drill-down capability through increasing levels
For the most part, financial consultants are familiar with the       of detail
organizing principles behind the menu hierarchy, but if a        q   keep information up to date
financial consultant is looking for information on a product     q   provide versioning and archiving of material for
he or she knew little about, finding the information can be a        regulatory compliance
frustrating experience.    For direct access to material, the    q   reduce paper
system provides keywords to the product chapter menus.           q   provide for searching of material both by attributes and
Individual documents can be referenced by supplying the              full-text retrieval
topic number and page number along with a keyword.               q   build an easy-to-use authoring environment
Keywords are limited to 14 characters and are difficult to       q   provide templates to create content in a WYSIWYG
remember. Financial consultants often have “cheat sheets”
                                                                     environment
pasted to their terminals to remember keywords to their          q   provide on-line versions of existing print material, such
favorite chapters.     Moreover, for a financial consultant
                                                                     as newsletters, client and financial consultant marketing
whose time is money, the menu system is unbearably slow.
                                                                     collateral and advertising
When a chapter is first added to the system it is created on a
                                                                 q   create and publish marketing information in multiple
                                                                     views depending on the twget audience (financial
staging file where it can be thoroughly reviewed by the
central editing staff and by the Legal and Compliance                consultants, clients or the public)
department.      Once a chapter has been moved to the
                                                                 q   create an automated workflow process to review
production file, future updates are immediate without                content
review.      Authors update the material through a CICS          q   enable stronger editorial and legal review of the
program which provides some minimal word-processing                  material even after initial publication
capabilities. No notification of the changes are issued either   q   provide bookmaking facilities for financial consultants
systematically or procedurally. All changes on the system        q   plan for integration with other applications and general
are archived for legal purposes, but there is no versioning of       system infrastructure
the documents.       There is no mechanism for restoring
material that has been erroneously written, forcing authors to
Note the similarity in these requirements to those specified    financial products have different     pieces of information and
for an industrial strength hypermedia        system for an      we could not generalize the          entities and relationships
engineering enterprise as reported in the classic paper by      between them. How then could         we go about designing an
[13]. Hypermedia as implemented on the Web quickly              unstructured repository and still    manage it in a systematic
suggested itself as a mechanism for delivering text, graphics   manner ?
and multimedia material in an integrated environment.
Hypertext would allow a less rigid structure for the            Since our primary requirements     were to author, store,
documents     and for direct     linkages   between related     manage, and deliver unstructured information and since we
documents. Related documents could point directly to each       could not strictly apply any of the above-mentioned
other and common material could be linked to from               hypermedia design methodologies, we decided to derive our
numerous places without duplication.                            own. Our methodology is largely based on RMM since one
                                                                of the authors was already familiar with it [1]. It is also
The traditional Web model of authoring and managing             based on some of the capabilities of the specific document
HTML documents on a Web server is insufficient for the          management system we chose for managing our information
requirements of the system. We expected the old system of       base.
40,000 plus screens to translate to about 10,000 hypertext
documents, In addition, we had to manage at least an equal      Web application development is similar to the development
number of brochures in PDF format and marketing material        of traditional/client-server   applications and it is iterative in
in other multimedia formats.       Given the requirements of    nature. Figure 1 shows a simplified view of seven different
managing     unstructured    documents,   assembling   views,   phases of our design and development approach. Following
providing restricted access to a diverse and distributed        sections will describe each of           them in greater detail,
authoring community, maintaining versions and enabling          Information      Architecture    ($2. 1)     involves    gathering
workflow, a tile system alone was inadequate for building       information, analyzing, organizing, and identifying patterns
such a system. A relational database was also not sut%cient     of hierarchies and relationships between various units of
due to the unstructured nature of the material. This has also   information.     User Inte$ace and Navigation Design ($2.2)
been the experience       of    Intermedia researchers   who    includes    designing      the presentation,    orientation    and
developed their own object-oriented database to overcome        navigation elements of the Web interface. Content Creation
the limitations of a relational database for hypermedia         and Authoring (!2.3) focuses on the identification               of
applications [15].                                              templates, selection of authoring tools, specification of
                                                                authoring guidelines, methods and procedures, and creation
A review of document management systems suggested that          of content. Workjlow and Document Management ($2.4)
one built on an object model would resolve some of our          addresses review, approval, management of content and
concerns.     It would allow us to store documents as           versions, and registration of attributes.      Publishing ($2.5)
components, assemble them into different views, provide         describes the generation of various views of documents for
privileged access to authors, and institute a workflow. In      preview, editorial and legal review, and final publication.
addition, hypermedia would allow us to create relationships     Document Review and Link Management ($2.6) presents a
between     documents.        Hypermedia    and   document      testing strategy for document and link integrity.          Finally,
management seemed to provide the appropriate solution to        Search and Retrieval ($2.7) explains how we identified fidl-
our problem, but at the time we started the project, no         text and attribute-based         search criteria    and defined
mechanisms were available to tie them together.          We     implementation mechanisms.
decided to build a hybrid system which would integrate the
features of hypermedia and document management so as to         2.1 Information Architecture
leverage the strengths of both.                                 Similar to requirements analysis and data and process
                                                                modeling phases in a traditional systems development
2 PSM WEB DEVELOPMENT METHODOLOGY                               Iifecycle, PSM Web development           required information
The lack of systematic design methodologies for developing      gathering, analysis, and modeling.         During this phase,
Web applications      has caused information management         information sources, categories, and relationships between
problems both in the Internet and intranet environments of      various pieces of information were identified. This phase is
large corporations.   We did not want to build a Web site in    similar to the Requirements Analysis, E-R Design and Slice
an ad-hoc manner with a group of “webmasters” authoring         Design phases of RMM [11]. We reviewed existing content
and maintaining the site. We reviewed existing hypermedia       base and its overall hierarchical structure.     We wanted to
design methodologies such as HDM [7], RMM [11] and              provide an information architecture that would allow an easy
00HDM [14]. All of them require the application domain          and intuitive way for financial consultants to traverse the
to be abstracted in the form of entities or classes and         information categories without being overwhelmed by the
relationships. Information units in our PSM domain were         information structure, that is, reduce cognitive overhead [4].
essentially unstructured but related to each other. That is,    We additionally required the ability to move between and
we could not arrive at a rigorous E-R model.        Different   across different levels of information without going up and
down a hierarchy.           The use of hypertext           was crucial in             level, a product has dimensions which provided links to
achieving this end.                                                                   other categories to which the fund might belong. In this
                                                                                      way, one could traverse the hierarchy across from the
In order to derive an information architecture, we chose                              ‘bottom’ to see other one-to-many relationships.      For
Mutual Funds as a sample area under a group of products                               example, a list of others funds belonging to the same
called Investments. At the highest level, there is descriptive                        category as Pacific Fund could be displayed.  Dimensions
information about Mutual Funds and they could be further                              are similar to indexes in RMM.
classified into Merrill Lynch Mutual Funds and Non-Merrill
Lynch Mutual Funds.                                                                   Not all products have the same organization of information.
                                                                                      For example, products such as equities, futures, and options



              I                                                         Integration
                                                                       Irsfrastrocture




              1’
                    e***e*+                                                                                                           I



                                       Figure 1: Phases of the PSM Web Development                    Methodology.




Merrill Lynch Mutual Funds are further classified based on                            may be organized differently with a totally different set of
their investment strategies such as Aggressive Growth,                                topics or components. This is exactly the reason why we
Growth and Income, Capital Appreciation, etc. A specific                              could not arrive at a generalized structure or an entity-
Mutual Fund, for example, Merrill Lynch Pacific Fund                                  relationship model. It must also be noted that we did not
belongs to one of these categories. At this lower most fund                           totally dispense with the hierarchical nature of information.
level, information about the fund can be broken down into                             As much as cross-references        are important to related
well-defined topics or components such as Description,                                products, hierarchies are equally important to show where
Client Benefits, Client Suitability, Financial Consultant                             the products belong in the overall schema.
Benefits,    Sales  Charges,    Performance,   Phone   List,
Enrollment, Marketing Material, Ideas, Competition, etc.                              The final output of this phase was an Information
Components are similar to slices in RMM [11]. Also, at this                           Architecture or Map showing structure (hierarchies) and
                                                                                      relationships (links).     Figure 2 shows the default area
                                                                                      hierarchy for a few products without any cross-links between
                                         Products                                     products or product groups. For example, hypertext links
                                                                                      could be provided from ML Fund D under Mutual Funds =>
                                    l-%---”                                           MLAM => Income to other mutuat funds categorized under
                              investments           Other Products
                                                                                      Non-MLAM        or to Other Products         based on some
                                   I
                             Mutsssl Funds                                            relationship. This initial architecture and an initial prototype
                                   I                                                  of the user interface were designed with the assistance of
                                                    [
                  MLAM                        Non-MLAM
                                                                                      Dynamic Diagrams, Inc., [6], an information architecture
                                                                                      and design firm. Content experts will be trained in such
                                                                                      information analysis and mapping techniques.
   Growth          lneome       Growth &lncome            Tax-Exempt

                                                                                      2.2 User Interface and Navigation Design
                                                                                      During this phase, we identified presentation, navigation,
     mkl:=!                                                                           and orientation components of the Web application. This
                                                                                      phase is a combination of two separate phases in R.MM,
                                                                                      namely, Navigation Design and User Interface Design [11 ].
        Figure 2: Example Product Area Hierarchy.
                                                                                      The Information Map identified earlier was prototype using
                                                                                      HTML in order to represent the structure and relationships.
Creating Web pages that contain elements of consistent                          The three-frame model contains the following:
navigation    and format      will better enable       financial
consultants to find and traverse the information. A standard                    .       A layout document defining a HTML                         frameset
look and feel will enhance the usability of the system and                              containing three frames: body, left and top.
enhance accessibility of the wealth of information that is
presented.     In order to produce this consistent and                          .       A body frame of text describing features of the product.
predictable appearance     of PSM Web pages, the user                                   The core marketing content of the product and its
interface was created using frames in HTML and content                                  various topics or components will be in this frame. This
was assembled out of components.       This approach allowed                            body document will be assembled           out of these
us to control the look and feel and also to provide                                     individual   components    as described     earlier. The
consistency    by systematically      imposing     navigational                         component     titles become named anchors           in the
elements on the assembled content. Otherwise, we would                                  assembled body document for linking purposes.
have to rely on content authors following interface standards
and guidelines which would have been difficult to enforce.                      .       A top frame for a global navigation tool bar, a locator
Although this was somewhat limiting in terms of allowing                                bar and title area. The global navigation toolbar allows
for individual creativity, we felt that the trade-off provided                          the financial consultant to navigate to landmarks such as
us with a more usable and functional system which was more                              Home, Subject Index, Glossary, Text Search, Product
critical.  The details of this component based publishing                               Scan, etc. The locator bar shows where one is in the
model will be presented in Section 2.3,                                                 default information hierarchy. Activation of any of the
                                                                                        links in the locator bar will result in traversal to higher-
We arrived at a three-frame based HTML presentation as                                  level information related to the currently displayed page
shown in Figure 3. While the top frame provides global                                  of information.
navigation cues [12], the left frame provides local navigation
cues [12] into the body frame that is composed of                               .       A left frame containing links to topics and to other
components. Frames also enable changing certain pieces of                               related areas. Component labels such as Description,
information without affecting others. For example, if the                               Client    Suitability, Client Benefits,  Performance,
information for a particular set of products is spread over a                           Financial Consultant Benefits, Sales Charges, What’s
number of Web documents and if a navigational component                                 New, Ideas etc., provide links to named anchors
is changed due to the addition of say some higher level                                 contained in the assembled body frame. This frame
information, editing each and every document to include the                             provides local navigation cues. It also has links ro
new navigational component would have involved a lot of                                 dimensions that enable traversal across hierarchical
maintenance. Instead, we preferred making a global change                               categories, showing other one to many relationships.
that is reflected across many documents.                                                For example, a fund that is by default categorized as
                                                                                        investing in the Pacific Rim could also be categorized




                             Giabcl
                            Na.lgatiom   _
                                Bar                                                                                                   Laubr
                                                                                                                                         Bar
                               TIUe

                                                                Mur$pslcin
                                                                MS Smd SMIU lmG-t@m
                                                                                  UC4tal -IMIM                o-       thmqh
                                                                                                                                  i
                                                                                                                                  2
                                                                                m of CUP@JOIU dwvcM            in far East-m or
                                                                m$tun   PJcil%3
                                                                              CUMIMS.                                                 IEcmipth
                                                                                                                             q
                                                                                                                                      Compaent
                                                                rho fuml tic  ~#usd       WM-   m th8sB camtrns but can also
                                                                  !& qiauiih ~tii95     thm!mh favor-           m
                                                                                                        CIWIWUY ovmunts




                                                           II                       -           end)
                                                                                         M,Wx(frmt
                                                                                    0    BxxW(back     em!)
                                                           II                       <    !cxxxx(lwel   load)
                                                                                    u    Wxxx( front   and)
                                                                \. . . . . ~~~~~~~~                                ~~~~~
                                                                U&t Suitubilily                                              4
                                                                                                                                         Cuu$
                                                                                                                                       Suhbuity
                               Butlona _
                                                                                                                                       Cmlpnrat




                                             Figure 3: Snapshot of the three-frame                     user interface.
                                                                   underlying relational tables. It enables the storage, import,




                                     a
                                                                   export, management, and retrieval of documents and other




              v
                                                                   ‘objects’ in various formats including HTML, PDF, MS
               U!lmid
                                                                   Word, etc. Authors, editors, legal reviewers, and a technical
                                                   web             team will all work with this DMS to create, store, manage,
                                                                   and update content. The DMS will serve as the “publishing
                                                                   engine” for the purposes of preview or review on to a
                                                                   Staging Web server and also for release to the Production
                                                                   Web server that will be accessed by financial consultants.
                                                                   The Content Creation and Authoring phase is explained in
                                                                   terms of components, attributes, links, views, templates, and
                                                                   tools.

                                                                   Components. A dedicated team of technical people (graphic
                                                                   designers,   webmasters,      application   programmers,    and
                                                                   document administrators)       will create standard interface
                                                                   elements such as navigation bars, images, and administer the
                                                                   DMS and Web servers as shown in Figure 4. The technical
     Figure 4: PSM Web Authoring and Publishing                    team will also establish access privileges and working
                    Architecture.                                  folders for authors, editors, and legal reviewers.      Content
                                                                   for the PSM Web will be created by authors in the form of
                                                                   components using templates and “checked into” the DMS
    as a Growth Fund.        Hence, the dimension called           under version control.          Components will make up a
    “Growth” will display a list of other funds which belong       compound      document       (a    powerful    facility  within
    to the Growth category.                                        Documentum) that will be displayed as the body frame as
                                                                   shown in Figure 5.
2.3 Content Creation and Authoring
After identifying the information architecture, the user
                                                                   AWxles.     Similar to the concept of object attributes [13],
interface and navigation designs, the systems development
                                                                   the DMS provides standard attributes such as Title,
team was challenged with identifying standard methods and
                                                                   Keywords, Author, Creation Date, Modified Date, Version
procedures and required tools for authors to create content,
                                                                   Number, etc., for all documents.      We also included the
This phase is similar to the “data entry” or data population
                                                                   concept of component-level attributes that are associated
phase in developing traditional database applications.   It is
                                                                   with the individual components and product-level attributes
dso similar to Conversion Protocol Design and Construction
                                                                   that are associated with the assembled body document.
phases of RMM [11].
                                                                   During the check-in process, authors identify component-
                                                                   level attributes.   For example, each component can be
One of our primary requirements was to enable subject
                                                                   assigned Boolean values on target audience attributes such
matter experts, currently creating marketing literature in the
                                                                   as Financial Consultant View, Client View, and Public
form of brochures, pamphlets, and prospectuses, to author
                                                                   View. Component-level attributes such as a component title
for the new medium without affecting their daily routines.
                                                                   will enable generating the topics in the left frame to named
They were spread in about 200 product marketing areas
                                                                   anchor sections in the body t%une. Product-level attibutes
throughout the company at various locations. We continued
                                                                   describe the characteristics of the product (meta-data) and
using Mutual Funds as an example for a pilot, Subject
                                                                   enable business criteria based information retrieval using a
matter experts from this product area were also available for
                                                                   search engine. Note that a specific product can have multi-
consultation regarding the content, its organization, and its
                                                                   valued attributes. For example, on the attribute called Client
presentation.
                                                                   Segments (consisting of demographic information), a mutual
Content authoring and publishing processes for the PSM             fund may satisfy the needs of one or more client segments.
Web are dependent on the features offered by a centrally
                                                                   Links. Financial consultants have concentrated on marketing
managed, commercially available enterprise-wide document
                                                                   the products they know best. The current emphasis is on
management system (DMS) called Documentuml.                  We
                                                                   promoting a client-focused approach by educating them on
refer the reader to Figure 4 for all discussions in this section
                                                                   how products are related to each other and how they could
through Section 2.7. Documentum is an object-oriented
                                                                   be packaged to satisfy a client’s specific investment needs.
system residing on top of a relational database, It provides
                                                                   This is where hyperlinking related products will be effective.
an abstraction layer that maps object-oriented constructs to
                                                                   This linking is at the discretion of the product marketing
                                                                   area as some would like to link to related products while
1 http://www.documentum.    com                                    others would like to stand alone for the sake of product
                                                                                                                                        Components
                                                                                                                          /..---             -------.-
                                                                                                                                   l—
                                                              Browser          Menubar                                                                   desc.htm
                                                                                               1
                                                                                                                                                         componmt.




                                       -h==
                   Hand.    crafled           Ho..      ,.           x
                                                              C4,[red,     .. ...... i                                                                   specifm
                   top.htm.                .“.,l... I, M.,,   ,,,.,’   I   ML*M                                                                          atrrlb.les
                   Contains       title,   Merrill Ly ch Pacific                  Fund
                   majorentry                                                             1.
                   w~inls. locator
                                                              Descriptl.      n                                                                          Csuit.htm
                   iar   and
                   global     navigation                                                                                                                 comfmn.mt.
                                                                                                                                                         specific
                   Generated                                                                                                                             attributes
                   Ieft.h tm.
                   Contains                                                              11
                   named       links
                                                              Client Suitability                       Assembled compound




                                                                                              I
                   from title                                                                                                                            cbene.htm
                                                                                                      -document body.btm,
                   attributes of                                                                                                                         compOnent-
                                                                                                       Contains product-
                   components                                                                                                                            specific
                                                                                                       specific at frihufes
                   and link,r 10                                                                                                                         attributes
                                                                                                       combined and stored in
                   Iayout.htm    of        ~.~~~~~=~                                                   META tag
                   d,menslond
                   pages
                                                                                                                                                         perf.htm
                                                                                                                                                         compome”t.
                                                                                                                                                         specific
                                       T                                                                                                                 attributes




                                                                                                                             ----------------

                                                   Figure 5: Component-besed                       authoring and publishing.


differentiation. The folder structure in the DMS and on the                                   with them that were available in the DMS. These templates
Web server was organized in such a fashion that authors                                       had irdine instructions as comments. Authoring guidelines
need not have to traverse more than two levels up or down in                                  and instructions were also prepared. While having universal
order to link to pages of other products.                                                     templates for all product areas is not possible, they could at
                                                                                              least be generalized across certain groups of products.
 Views. While the primary users of our content will be
financial consultants, a smaller subset of the content will                                   Tools. Although most Web authoring tools have evolved
also be delivered to clients (through a dedicated online                                      rapidly over the past months, they are still not ready for the
service) and a much smaller subset to the general public                                      novices that represent our authoring community. We agree
(through the public Web site).         Components such as                                     with [2] about the observation             that the authoring
Description, Client Benefits, and Client Suitability can be                                   environment    for the WWW is analogous to second-
delivered to all three audiences.       On the other hand,                                    generation programming       languages and that we need
components such as Financial Consultant Benefits, Phone                                       facilities  equivalent   to fourth       generation   languages.
List, Ordering, Competition,     and Marketing Ideas are                                      However, we expect that these tools will be enhanced and
specific to financial consultants only.     Component-level                                   that there will be better products in the near future.
attributes enable the re-use of components by directing the                                   Breaking up the material into components and topics has
assembly of the appropriate       ones for different target                                   made it easier for authors to “fill in the blanks” that make up
audiences.    Figure 6 shows how different components                                         their portion of the final composed document.          We have
having different component-level attributes are assembled to                                  currently chosen a HTML authoring tool that is an add-on to
produce three different views. For example, all components                                    a commonly used word processing package to reduce the
having the Financial Consultant (FC) View attribute turned                                    learning curve for authors.
on will be assembled together to form the FC View of the
product.                                                                                      2.4 Workflow and Document Management
                                                                                              This phase is unique to the development of our system.
Temp/ates. Based on the information architecture and the                                      There is no parallel in RMM although a storage and
components that we identified for Mutual Funds, we                                            management system is implied [11]. Here, we describe the
established rempfafes required for the products within this                                   organization o~~okfers and other features of the DMS such
category and the intermediate    levels of documents (or                                      as compound documents,          access control, concurrency
pathways) to be traversed. Templates allow the system                                         control, version control, attribute registration and workjlow.
design team to focus on structure and presentation while
encouraging non-HTML savvy authors to concentrate on                                          Organization of Folders. The DMS holds sets of documents
issues of content.    Each of the topics or components                                        in ‘folders’ for each individual product. Each product folder
mentioned earlier had blank HTML templates associated                                         contains the sub-folders: Components, HTML, Images,
                             lhmlp. mc.t
                                                                                                    Pmd.<t      Attribute.
                             Attrib”lc,         C.nqmnents                             FC View




                                                                            J
                               ~c                                              D_ri@”



                                           l-l
                                               Duc,lptl..
                                                                               Client Suitability   Next G,..,.,,..
                               C Ii..t         ML P,,, (,c Fund
                               F’ublk          15                              Clint    B,, mtl
                                                                               P., form.llc.        Educalmn




                                           1                      I




                               m
                               Clint


                                                                          W==-l

                                         Figure 6: Views assembled from reusable         components.



Audio, Video and PDF. The HTML folder will always                     these built-in features,              concurrency      control   and version
contain the frameset layout top, left and body frames. The            control are achieved.
Components folder across all products will always contain
documents with the same names.             This allows for a          Attribute Registration.    The DMS provides interfaces that
generalized naming convention across all products and                 can be customized for the on-line entry of document
simplifies automation. The organization of folders does not           characteristics or attributes as part of its functionality. This
go beyond a two tier structure. Organizing folders into a flat        facility allows the authors to register component-level
hierarchy facilitates the creation of relative links during the       attributes and product managers to register product-level
authoring process.                                                    attributes. All documents stored in the DMS have default
                                                                      system attributes assigned to them such as Title, Keywords,
Compound Documents. Our design makes extensive use of                 Modified Date, Authors, etc.
a feature of the DMS known as the Virtual Document
Management (VDM) facility. Using the VDM we can define                Workflow. We envision three different roles witbin the
compound documents composed of or associated with                     product marketing areas. As shown in Figure 4, there will
components. Components can exist as actual documents or               be a set of authors who create content based on their
as symbolic ‘links’ or ‘pointers’ to other documents in the           knowledge of a product, a set of editors who oversee the
DMS. The body document that is a part of the frarneset                content and graphical standards, and a set of legal reviewers
making up the presentation        is actually a compound              to ensure regulatory compliance.     These three groups of
document.   Components are written together as one body               people will be assisted by a central technical team of
document directly to the Web server from the VDM utility.             developers   who will be responsible       for the smooth
                                                                      functioning of the DMS and the Web servers, establish
Access Control. There can be multiple authors per product             database comections to real-time information and maintain
and access is restricted only to those documents that they are        Web link integrity.
entitled to edit. For example, user interface elements are out
of reach of the authors.      The DMS also supports group             Authors can preview their work in progress by triggering a
privileges which can be tine-tuned.                                   process which assembles the components in the DMS based
                                                                      on target audience attributes and publishes the HTML
Check-in/Check-out, Concurrency Control, Version Control.             documents to the appropriate Staging Web server directory.
The DMS supports full version control. This provides for              Authors use a standard Web browser to view the final
maintaining a ‘current version’ and all prior versions of             composition and can continue working on content till they
documents as desired. Any components to be authored are               are satisfied. They can then forward a notification to editors
first ‘checked-out’ of the DMS for editing. When an author            for review.     The editor has the privileges to change the
‘check-out’ a lock is placed on the document and it belongs           component content and re-publish the pages. The author is
only to the author. Once saved, a document is ‘checked-in’            informed of the changes and can review the re-assembled
and a new version can be created or the current one                   document on the Staging Web server to see if they approve
overwritten. The compound body documents as well as the               the changes. The editor routes the components and the
individual components can have various versions.       With           location (URL) of the assembled document to legal
reviewers.    Assuming that legal reviewers are satisfied,      2.7 Search and Retrieval
editors inform authors that all is well and the authors can     This phase is also similar to Testing in RMM where
finally approve the assembled document for release. At any      documents are accessed by applying search criteria. Before
point in this workflow,          participants  can exchange     being published to the Staging and Production Servers, the
information by adding to the workflow package that is sent      assembled body document will be assigned meta-data such
along with the documents. Note that, the workflow between       as client segments, financial goals, investment objectives,
authors, editors, and legal reviewers can iterate a number of   etc. These attributes or meta-data will be in the DMS. On
times. Periodic notifications will be sent to authors to re-    publishing, the attributes will be written out inside a META
examine content for currency.                                   tag of the body document,

2.5 Publishing                                                  As shown in Figure 4, both Web servers will be indexed as
This phase is similar to Runtime Behavior Design and            content is updated using a commercially available full-text
Construction phases in RMM. The process of publishing is        search engine2. In addition to full-text searches, the search
automated so that any time a component is modified, the         engine also has a facility called region-based searching
compound documents it is used in will be re-assembled and       which allows for attribute-based searches. Through search
re-published thus ensuring currency of content. Server side     forms, financial consultants can now retrieve products that
programs are written for the DMS to periodically assemble       satisfy specific business criteria such as Client Segments,
and publish updated documents.        The HTML documents        Investment Objectives or Financial Goals.
comprising the frameset layout (top, left, and body) for a
particular product along with its related Images folder and     3 RELATED WORK
other media folders will be published for preview or review     Requirements for the PSM Web were a lot similar to the
to a specific directory on the Staging Web server as shown      requirements for industrird strength hypermedia systems
 in Figure 4. The left frame will be created from the Title     suggested by [9, 13]. We too had requirements such as
 attributes of the individual components and other relevant     interoperability,   object attributes, interactive  authoring,
dimensions such as Fund Manager, Type of Fund, etc. The         templates, navigational aids, access control, version control,
 body document will be assembled out of the components          query mechanisms, composites, concurrency control, and
 that have like target audience attributes.    The directory    collaboration.     Documentum     helped us to successfully
structure on the Staging and Production Web servers will        address all these requirements to a large extent. While some
mirror the folder structure within the DMS.         Authors     researchers     [16] have suggested modi~ing       HTML to
perform publish for preview to preview the content created      support versioning, we achieved it through the built-in
 by them. Editors perform publish for review after making       capabilities of the DMS where version information is stored
necessary changes and informing the authors, Authors in         as an attribute.
conjunction with the document management administrator
perform publish for release after the document has been         A lot of the ideas for our development methodology were
approved for release to the financial consultants or to the     derived from RMM [11].          The table on the next page
other two audiences. The approved documents and related         compares our design and development phases with those of
media objects will be placed on the Production Web server       RMM. Note that some of the phases in RMM are split
which will be accessed by financial consultants. The Client     across some of our PSM Web development phases.             In
and Public views will be published to other Web servers         RMM, entities have attributes and slices are made of groups
residing outside the corporate firewall.                        of attributes. In our case, both components or slices as well
                                                                as assembled compound documents have attributes.
2.6 Document Review and Link Management
This phase is similar to Testing in RMM where documents         4 DESIGN TRADE-OFFS AND TECHNOLOGY LIMITATIONS
are accessed and links are traversed. Documents published       In designing and developing the PSM Web site, our
to the Staging and Production        Web servers will be        emphasis was to integrate off-the-shelf products as much as
periodically checked by administrators.  Also, as shown in      possible and leverage existing technology. We had to accept
Figure 4, a link checker or a Web walker such as                some of the limitations of current software resulting in
WEBanalyzer will be executed regularly on these two Web         compromises     on our design.          For example,  current
servers in order to check for missing links. The idea of a      generation Web authoring tools do not support several of the
link service has existed since the days of Intermedia [17].     features pointed out by [8] such as composites, version
We believe that in order to ensure 100’% link integrity right   control, extensibility, and collaboration. Due to the infancy
from the time of creation, relationships between HTML           of WYSIWYG HTML authoring tools, we could not provide
documents and dependencies between HTML documents               for “iron-clad” templates as requested by the business team.
and other object types (images, audio, video) must be           The missing hypermedia functionality in the Web authoring
recorded permanently as soon as an author creates these
documents and establishes the links.                            2 For the sake of maintaining confidentiality agreement with
                                                                the vendor, we will not mention the name of the product.
environment can be augmented by using a DMS which                                        management “after-the-fact”, we would like to have a link
supports the features identified by [8] and [13]. Although                               management facility that could be fully integrated with the
Documentum has its own strengths, it has no understanding                                authoring tool and the document management system. Such
of relationships between documents created either as HREFs                               a facility would ensure link integrity during the authoring
or as references to objects such as images, audio or video.                              phase. This is similar to the concept of a link management
We carry out link verification “after-the-fact” using off-the-                           and delivery service for the WWW as suggested by [3]. A
shelf     Web walkers which report on relationships and                                  good link database will not only ensure 100% link integrity
dangling links.                                                                          but also facilitate visualization of the information space,
                                                                                         enable link change notifications, and capture attributes on
We also chose a static publishing model. That is, HTML                                   links (such as link traversal privileges).
documents are not generated “on the fly”, but published
when content is updated. We adopted this approach because                                5 CONCLUSIONS
most of the PSM information            is essentially  static                            Although the Web has simplified information  delivery, tasks
(description of a Mutual Fund does not change every day).                                such as building the authoring environment,        managing
Changing information      such as Performance      and Sales                             documents, and relationships between them are not easy to
Charges is presented by interfacing with applications that                               accomplish without a large technical staff. Corporations
produce live data feeds. This model may change if the                                    building large Web applications with similar requirements
newly introduced Web-based model of Documentum can                                       will be challenged with the same problems we encountered.
scale to the volume of transactions we anticipate when the                               We have successfully addressed major issues by adopting a
system goes live. In such a situation, documents and views                               systematic design methodology and by integrating document
can be generated dynamically based on access and link                                    management and Web technologies.       We believe that our
traversal privileges and queries.                                                        methodology can successfully be used to systematically
                                                                                         design and develop large-scale hypermedia applications.
Integrating the WWW with a number of tools to build
hypermedia applications has also been suggested by [5].                                  We have defined a solution where subject matter experts not
While a lot of the requirements specified by [13] can be                                 well-versed in HTML can create media-rich content and
addressed by integrating a wide variety of tools, we still do                            submit it for editorial and legal review electronically.  The
not have a totally integrated solution to create and manage                              reduction in cycle-time due to online notification and review
large hypermedia collections delivered via the Web. While                                needs to be measured.      The usability and validity of this
relational databases can be used to address link management                              approach has been successfully tested on a small-scale to
issues, they do not handle document management features                                  create content for one product area with the cooperation of
very well. Conversely, document management systems are                                   non-HTML savvy subject matter experts. We are currently
not good at handling relationships      between documents.                               in the process of customizing the authoring interface and
Therefore, we are exploring link management as an integral                               integrating a standard WYSIWYG HTML authoring tool;
part of the commercial document management system we                                     usability of such an authoring interface will be collected
have chosen for our application. Although, we achieve link                               only after the initial implementation phase in other product


   PSM Web Development Methodology                                                   ) Relationship Management Methodology [11]
   Information Architecture                                                           Ent@-Refatiarrship     Design and Slice Design
   Identifyinformationsources,categories,topics(components) and                       Identifyentities,relationships,slicesof entities- structuredinformation.
   relationships   - unstructured   information.
   User Interface and Navi@ion Design                                                 Navigation Design and User Irrferjiie Design
   Identify presentation elements, Iocaf and globafnavigation elements,               Identify accesspathsthat enable hypertext navigation between entities
   orientation components, and grsphicat elements.                                    and theirrelationships.Identifyindexesand guidedtours.

   Content Geafiorr and Authoring                                                     Conversion Protocol Design and Construction
   Identifycomponents, templates,authoringguidelines,attributes,views                 Devise a set of rules (or programs) to convert design elements into
   and tools to create content.                                                       physicat objects in the target hypermedia application. Populate the
                                                                                      database with content.
   Workjlbrv and Document Management
   Identify storage, version control, workflow, document review, access
   control, concurrency control mechanisms.
   Publishing                                                                         Runtime Behavior Design and Construction
   Assemble and publish different views for different audiences.                      Design and develop programs that will control how the application
                                                                                      generates and retrieves information, and define interaction behavior.
  Document Review and Link Management                                                 Testing
  Check document access and links between documents.                                  Access documents, navigate through links.
  Search and Retrieval                                                                Testing
  Full-text   index HTML documents         and META tar reeions   for attrihrrtes.    Search frrrdocuments
marketing areas.   The pilot approach was also taken to           4.   Conklin, J. E. (1987). Hypertext: a Survey and
provide subject matter experts with some exposure to new               Introduction.   IEEE Computer 20(9), 17-41.
ways of thinking and creating content for the new medium.
                                                                  5.   Dieberger, A. (1996). Browsing the WWW by
We have provided a common infi-astructure where content                Interacting with a Textual Virtual Environment - A
can potentially be authored once, but presented to multiple            Framework for Experimenting with Navigational
audiences (financial consultants, clients, and the general             Metaphors. Proceedings of Hypertext’96, 170-179.
public) by adopting the component based publishing model.
We have provided templates, tools, and guidelines for             6.   Dynamic Diagrams, Inc.
authors to create, update, and publish content. We have also           http://www.dynamicdiagrarns.com.
addressed the issue of regulatory requirements of version
control and archiving of individual components       as well as   7.   Garzotto, F., Paolini, P., & Schwabe, D. (1993). HDM
assembled documents. Information presented in the form of              - A Model-based Approach to Hypermedia Application
components or topics is an easy transition from the current            Design. ACM Transactions on Information Systems,
system. We have linked HTML documents to related                       11(1), 1-26.
marketing material such as glossy brochures in PDF, audio,
and video presentations.        We have increased a financial     8.   Halasz, F.G. (1988). Reflections on NoteCards: Seven
consultant’s     accessibility  to information   by not only           Issues for the Next Generation of Hypermedia Systems.
providing    an intuitive      user interface  and navigation          Communications of the ACM 31,7, 836-855.
mechanism but also providing attribute-based and full-text
retrieval facilities. We have simplified link management by       9.   Halasz, F.G. (199 1). “Seven Issues” Revisited.
specifying rules as to what kinds of documents to which                Hypertext ’91 Keynote Address.
authors can provide links.         We have also provided a             http://www.parc. xerox.corn/spl/projects/halasz-
feedback facility for financial consultants        to directly         keynotel.
communicate with product marketing areas. We are in the
process of providing personalized, server-side bookmarks.         10. Hopper, S., Hambrose, H., & Kanevsky, P. (1996).
Prelimimwy usability studies with financial consultants have           Real World Design in the Corporate Environment:
shown that they “like” the user interface.      Many of them           Designing an Interface for the Technically Challenged.
wanted to know “when can we have it ?’                                 Proceedings of CHI ’96,489-495.

6 ACKNOWLEDGMENTS                                                 11. Isakowitz, T., Stohr, E., & Balasubramanian, P. (1995).
This project is a result of the combined efforts of the PSM           RMM: A Methodology for Structuring Hypermedia
Web Team and we would like to thank the following: Mike               Design. Communications of the ACM, 38(8), 34-44.
Snizek, Phil Gilligan, Lorraine Franza, Gail Davala, Luanne
Arico, Joe DeFranco, Ray Walters, Marc Harbatkin, Emma            12. Kahn, P. (1995). Visual Cues for Local and Global
Jaffe, Rich Caran, Robert Raud, Susan Hopper, Melenda                 Coherence in the WWW. Communications of the
Moore, Gururajan Rae, Bruce Weimer, Piyush Pandya,                    ACM, 38(8), 67-69.
Oliver Smith, Paul Kahn, and the team at Dynamic
Diagrams, Inc. We would also like to thank Paul Kahn for          13. Malcolm, K.C., Poltrock, S.E., & Schuler, D. (1991).
his comments on the initial draft.                                    Industrial Strength Hypermedia: Requirements for a
                                                                      Large Engineering Enterprise. proceedings of
REFERENCES                                                            Hypertext ’91, 13-24.
1.   Balasubramanian, V., Ma B., & Yoo, J. (1995). A
     Systematic Approach to Designing a WWW                       14. Schwabe, D., Rossi, G., Barbosa, S.D.J. (1996).
     Application, Communications of the ACM, 38(8), 47-               Systematic Hypermedia Application Design with
     48.                                                              00HDM.     Proceedings of Hypertext ’96, 116-128.

2.   Bieber, M., VitaIi, F., Ashman, H., Balasubramanian,         15. Smith, K.E., & Zdonik,S.B. (1987). Intermedia: A
     V., & Oinas-Kukkonen, H. (1997). Fourth Generation               Case Study of the Differences Between Relational and
     Hypermedia: Some Missing Links for the World Wide                Object-Oriented Database Systems. 00PSLA ’87
     Web. Forthcoming in the International Journal of                 Proceedings, 452-465.
     Human-Computer Studies.
                                                                  16. Vitali, F., & Durand, D. (1995). Using Versioning to
3.   Carr, L., Davis, H., De Roure, D., Hall, W., & Hill, G.          Provide Collaboration on the WWW. Proceedings of
     (1996). Open Information Services. Proceedings of the            the Fourth WWW Conference. http://www.w3.orgl
     Fifth WWW Conference, http:lldiana.ecs. soton.ac.ukl             pub/WWW/Joumal/l/vitali.    190/paper/190.html,
     -lac/WWW96/Overview.      html.
17. Yankelovich, N., Haan, B., Meyrowitz, N., & Drucker,
    S. (1988). Intermedia: The concept and the
    construction of a seamless information environment.
    IEEE Computer, 211, 81-96.