Relationships provide an interesting segue into indexing.In view of the discussion above concerning Concept/Relationship interplay, it is necessary to consider the impact this would have on index construction.Since the same Concept supports Topic and Category, each a Relationship, it is the Relationship that would determine how to separate the two into topical and categorical indexes.Alternatively, they could be distinguished in the same index, drawing attention to the distinction. This simple example parallels MARC (650 and 655); other cases are likely to be less predictable. The 'degree' attribute (e.g. primary/secondary) supports an important distinction, especially in cases of high postings. This section's goal is to introduce issues, rather than provide solutions.More study is needed.
XOBIS provides a structured framework for many kinds of information. While this structure is intended to support improved indexing due to improved organization, it is not prescriptive regarding how derivative indexes should be built.The Principal Elements provide a point of departure. Regardless of how the pie was cut, troubling divisions always resulted. To alleviate problem categories that fall on the "wrong" side of a cusp vis à ¶is a particular index, overlap could be defined.Since automating, Lane Medical Library has included conference names in both the name and title indexes for this reason.When librarians have difficulty deciding where to assign a particular value, it is likely that users will have difficulty in choosing the "right" index.Some of the Principal Elements may be expansive enough to split; others might be lumped. A third option would be to include subsets from one or more of these in a various cross-Principal Element indexes. These examples suggest some possibilities with potential subsets listed under an index and their Principal Element of origin on the right.
| Organization Index | |
| Family names | Name |
| Organization names | Organization |
| Jurisdictions | Place |
| Title Index (Entry/Varia) | |
| Titles of works | Work |
| Meeting names | Event |
| Object names | Object |
| Language Index | |
| Computer languages | Work |
| Languages | Language |
| Markup Languages ? | Work |
| Scripts | Work |
| Keyword Index | |
| Words and phrases | String |
| Acronyms, codes, etc.* | <other Principal Elements> |
* Cf. the String element, Entry Substitutes and Varia for the distinction.
We envision another type of browse index, a structured, smart or value-added one, building upon hierarchical relationships indicated in various places in the foregoing text.Due to the uniformity of structure, these could be optimized for each Principal Element. A topical index could display subdivisions and relationships as optional substructure, eliminating clutter from a purely alphabetical listing.Similarly, geographic and organizational substructure could be displayed hierarchically in multiple tiers.Personal names present interesting possibilities.Both forename and surname entries could be included in the same index, distinguished from and subarranged by their reciprocal. Titles may offer the most promise in this regard. Series or periodical titles could offer up their relationships, especially subordinate ones such as articles and chapters in highly organized structures.Time also offers intriguing possibilities and challenges. The same hierarchical structure would allow "drilling down" chronologically, but provision for "bursting" ranges of values and considering differences of usage by relationship need to be considered.The basic idea is to hide substructure until the user exhibits interest by choosing particular value.Navigational Relationships above provides the underpinnings for this indexing functionality.
Indexing is a complex issue. As a temporary way to test these ideas, we will consider building separate XML documents containing index entries that could be updated by editing software when XOBIS records change. Other indexing options may prove more effective. In any case, libraries should no longer settle for raw alphabetical listings. We can do better.Our goal should be to move this functionality onto the Web more effectively. We hope that XOBIS might play a role in this endeavor.
Another important issue in indexing relates to language.Our temporarily aborted Clustering technique is discussed at the end of the Language section.Current features in XOBIS supporting language distinctions need testing and broader input before deciding upon the best technique to support these robustly. Tillett and international community have pioneered in this area (59).We are encouraged that flexible solutions are possible to improve user options and index coordination and clarity.