Untangle the Web.

The proliferation of data boggles the mind, but an enterprise portal can deal effectively with its structured and unstructured sources.

The world's computers currently contain approximately 2.5 billion terabytes of data, and more is being added daily. The amount of information in corporate relational databases is a small fraction of this, but it is still a staggering 50 million gigabytes. Increasingly, success in the information economy will depend upon a company's ability to retrieve, correlate, package and customize this internal and external data for employees, business partners and customers.

Consequently, businesses are scrambling to build enterprise portals, and more than two-thirds expect to have one in place within two years. With so much riding on these enabling platforms, understanding and implementing them correctly from the ground up is important. The technology has to be flexible, scalable, and easy to administer and use.

As enterprises have added specialized databases, their users have acquired multiple digital persona. Each big enterprise application has to be accessed separately, and correlating the information retrieved from each is a largely manual process. When a new employee joins the company, setting up separate accounts for each enterprise application and providing even rudimentary training can take a couple of weeks.

An enterprise portal eliminates all this by providing a single point of access through a familiar Web interface, making the entire universe of information look like a single database. It understands the relationship that each user has with each underlying data source, and packages and presents information according to individually defined preferences.

A true enterprise portal must be able to deal with all the disparate structured and unstructured data sources in the extended enterprise. Traditional corporate databases tend to be highly structured, but increasing amounts of critical information are contained in semistructured and unstructured data stores, such as e-mail messages, word-processing documents, spreadsheets and HTML files. There is a wealth of information in these data sources that needs to be leveraged and delivered to the right people at the right time--and that is what the right enterprise portal can do.

CLASSIFICATION IS THE KEY

If all this information is to be retrieved, it must be classified appropriately, either manually or automatically. In the manual method, people look at each individual document and assign it to one or more categories. This is how some of the public information portals--like Yahoo--classify information. This labor-intensive process is expensive, however, and the classifiers are not always subject-matter experts. Also, both the content and its location become obsolete over time as taxonomies and business processes evolve.

With the automatic approach, administraters basically train the computer to categorize data accordingly. New documents are matched with similar ones. The computer, however, is looking for statistical patterns and does not understand the subtleties of human language. It may not know that "car" and "automobile" are the same, or how to determine which of 20 possible meanings the word "charge" is supposed to have. Similarly, the machine must be able to retrieve information based upon inexact or misspelled requests.

While computers are getting better at this kind of "machine learning," the most state-of-the-art machine intelligence available can still achieve no more than about 85% classification accuracy. Also, an additional 15% of accuracy is lost at each subclassification level, so completely automated systems tend to be limited to a couple of levels.

The best approach is to combine the manual and automatic methods, using humans to create the categories and provide some classification hints before turning the machines loose. With such a hybrid portal technology, you can take your existing document hierarchies--directories and subdirectories--and mirror them into the portal infrastructure. Any new document categories can be added as needed. PERSONALIZATION AND SECURITY

Data classification is only the first step, however. Different departments and individual users have different needs, and an enterprise portal must be able to distinguish among them and present information accordingly. It must automatically recognize you when you log on, and--based on your job function, access privileges and self-defined preferences--dish up the screens you want.

While all users are working from the same aggregated database, each should see only the relevant subset. All other information should be completely invisible. Some portal engines, however, list the entire results of a search, and then deny access to some of them. This frustrates honest users and provides hackers with a roadmap. A good portal engine can use access privileges to filter search results before presenting a document list to users.

Within these subsets, the portal technology should enable individual users to customize the organization and presentation of information. While the computer does not have full comprehension of human language, the query interface should make it appear so to the users as they interact with it.

Ultimately, enterprise portals will provide implicit personalization based on observation of individual user behavior. If you are constantly accessing certain types of documents, the portal engine will automatically find and point at others that might help you. It might also identify other individuals in your organization that are using the same types of documents, and send you to these experts.

Imagine being a Big Four accounting firm with teams of people working on various consulting projects. Productivity could be increased considerably if a team in one industry segment could leverage the expertise, documents, templates and processes that have already been developed and used by a team in a similar industry. An enterprise portal can identify and deliver these documents and enable the new team to spend less time searching and more time acting on it, thus boosting profits by a third.

Even relatively small companies with a few hundred employees can have serveral million documents that are accessible from their portals. The ratio of this Web-facing information to data that is still purely internal is about 1:20.

PLAN FOR GROWTH

An enterprise portal must be able to scale to handle multiple terabytes of different types of data scattered across different locations that are connected by different access methods. The engine must search through all the data sources, find the relevant information, and serve it up on demand to people around the world in unpredictable ways, all the while respecting individual access privileges.

The amount and location of data is just one parameter of scalability. An enterprise portal must also be able to scale to handle many users hitting it simultaneously. As companies get bigger and more global, their portals have to provide information around the clock and in different languages. Increasingly, enterprise portals have to support extranets of suppliers and customers, whose information access needs may be different from those of employees.

Customer-facing portals tend to focus on marketing, with the goal of stimulating transactions. While these applications typically do not involve tremendous amounts of data, they frequently have to draw upon multiple data sources before creating a personalized view. They also need to collect information about customer behavior and use it to make personalized offers of products or services. Customer-facing portals have to be able to handle peak loads triggered by promotional offers, or even by external events that are beyond the company's control.

Business survival in the 21st century may depend on implementing the fight enterprise portal technology. You need an enterprise portal because of all the complexity and diversity and hard-to-reach data throughout your company. You need technology that can grow with your data and internal and external user base, and evolve along with constantly changing business processes and market conditions. The ideal enterprise portal should combine the power of computerized classification with the subtleties of human intelligence to produce a knowledge-retrieval platform that will enable employees, partners and customers to work together efficiently in the increasingly virtual business environment of the future.

www.verity.com

Circle 252 for more information from Verity

Six steps for managing portal content

by Claude Vogel

All portal developers face the same problem: ensuring that content--primarily unstructured information--is clearly organized into categories the audience can understand. Although this is a complex problem, six basic steps start you in the right direction.

1 Know your mission. Write down what information sources you plan to index and who the users will be. You also need to understand how often the information will change and what kinds of updates are expected.

2 Understand your target audience. Make logical groups of the different types of people who will use your portal. For example, a medical portal might include doctors, nurses, patients, researchers and drug companies. Consider what these different groups need in terms of content, how much they already know about your subject, and how much expertise they bring to the task of finding and working with the various types of information resources you provide on your site.

3 Create an indexing structure. At this stage, you must think like a librarian to define a standard vocabulary for filing your content. You must create meta content to associate each article with an appropriate category or categories. Your content probably already comes with a creation date and title, but a useful index requires topic references.

4 Develop a categorization structure that is appropriate for your target audiences. This may require developing a different categorization structure for each group of users, depending on how significantly their needs and backgrounds differ.

5 Create a quality plan. You need to establish metrics to make sure that your indexing and categorization structures are offering appropriate levels of quality.

6 Develop a content-maintenance plan. A good portal is dynamic. As your site changes, you need a plan for managing the content and maintaining quality over time.

Vogel is chief technology officer at Semio, San Mateo, CA.

www.semio.com

Circle 251 for more information from Semio