About Relational Taxonomies

The use of multiple taxonomies to categorize the same information is implemented as the relational taxonomies feature of parametric search. Using relational taxonomies, end users can simultaneously navigate more than one taxonomy, drilling down and navigating to the information they seek in the manner most intuitive to them.

Relational taxonomies are useful in the common situation where a document can logically belong to multiple information structures, or where a business has multiple organizational perspectives for classifying the same information. For example, to help a user find a document about software companies in Sunnyvale, which of the following taxonomies should an application provide?

Drill down by location. Follow links from USA to California to Sunnyvale, hoping to find a Software Companies category.

 

Drill down by business category. Follow links from Businesses to Computers to Software, hoping to find a Sunnyvale category.

 

With relational taxonomies, the user is not bound to the taxonomy structure envisioned by the taxonomy creator. Instead, users can drill down and navigate as they wish, jumping from taxonomy to taxonomy wherever appropriate.

Consider a taxonomy showing the worldwide locations of car manufacturing plants. Figure 3-6 illustrates a portion of that taxonomy.


Figure 3-6    A taxonomy of automobile manufacturing plants



In this taxonomy, automobile manufacturing is organized geographically. Following one path shows that Ford Explorers are manufactured in (among other places) Windsor, Ontario in Canada. Comparing this taxonomy with the one presented earlier in Figure 3-5 shows that a relationship clearly exists between the two taxonomies. Both can be followed to arrive at the same document.

Each taxonomy has a static navigation path. However, when two or more taxonomies share overlapping sets of documents (such as documents describing the Ford Explorer in this example), they have a relationship that allows the trees, conceptually, to be traversed dynamically. This dynamic traversal effect occurs because selections from one taxonomy are dependent on the current selections from the other taxonomies.

Consider this traversal path through the two taxonomies:

1. The user starts the traversal from the root (Automobile models) of Figure 3-5, then selects the manufacturer Ford. Only Ford-related documents are selected.

2. The user then jumps to the Windsor link in the taxonomy of Figure 3-6. Now, only documents related to Ford vehicles manufactured in Windsor, Ontario are selected.

3. The user navigates to and selects the Explorer link, obtaining a list of documents related to Explorer manufacturing in Windsor.

Conversely, the traversal could follow the opposite path:

1. The user starts the traversal from the root (Automobile factories) of Figure 3-6, then follows the links down to Windsor. Only documents related to vehicles manufactured in Windsor, Ontario are selected.

2. The user then jumps to the Ford link in the taxonomy of Figure 3-5. Now, only documents related to Ford vehicles manufactured in Windsor, Ontario are selected.

3. The user follows that taxonomy down through Ford trucks to the Explorer link, obtaining a list of documents related to Explorer manufacturing in Windsor.

For an example of the interface an application might use to present relational taxonomies to the user, see Relational Taxonomies.