1. Fuzzy and incomplete information

The databases for this Encyclopedia and the Yearbook of International Organizations are maintained through constant re-editing in the light of new information from international organizations. The entities described may be organizations, world problems, strategies, values, or concepts. The conventional approach of "entity equals record" of course applies in the case of entities which are well-defined, stable and unambiguously named. In many cases the UIA is however dealing with fuzzy or incomplete information, only vaguely related to clear subject categories. It is up to the editors to identify an entity by a naming process which may only be provisional. In this sense entities have varying degrees of "stability".

2. Editing evolving networks

The real challenge lies in editing evolving networks of entities, especially where material in one entity record may have to be moved to another. Networks may evolve through the emergence of new challenges, through acquisition of new information or through new insights into how those networks can be more usefully ordered. In this sense the evolution is a reflection of a learning process, both of the international community in highlighting concerns and of the editorial process itself. In the case of world problems, for example, an entity may be named by a collection of from 1 to 10 phrases, each of which captures aspects of it. But the editing process, in response to new information or insights, may call for some of those naming phrases to be split off into a new problem entity -- possibly a broader problem, a narrower problem, or a sister problem. Portions of the descriptive text may also need to be moved. Corresponding to this splitting operation, there is of course a merging operation by which entities are fused together. The pattern of cross-references between entities may also be modified in a similar manner. These processes are summarized in Figure 1 (for entries) and Figure 2 (for relationships).

3. "Live" cross-references

Control of the actual movement of text is of course a standard feature of any text processor. These facilities have already been augmented by allowing the UIA editors to embed "live" cross-references in the text. Thus an entity's descriptive text may have many links to other entities (possibly in other databases) present as editable "text". Any changes in the cross-references are updated in the distant entities when the text is saved (as in any relational database).

4. Facilitating more perceptive editing

The UIA is concerned to develop its software framework to guide the editor in making more perceptive judgements about how the network can best be refined in the light of new information. These judgements involve a sense of hierarchy (broader, narrower) and a sense of functional relationships (aggravated, reduced by, etc) between entities. Hierarchies may not be simple tree structures. An entity may be grouped under several broader categories. Special problems include avoiding redundant links and inserting links to previously isolated entities which may not be readily identifiable in the database through any keywords in common with the portion of the network on which the editor is working.

5. Editorial experiments

Working under conditions of incomplete and fuzzy information, editors can be confronted with information on what appears to be a new class of entity. In the case of a "world problem", for example, this is usually signalled by a new class of "negative operator" defining the problematic nature of the problem. There may be a whole class of problems based on an existing operator such as "unethical" (eg "unethical physicians", etc). But a possible new operator such as "incompetent" may suggest a parallel class (eg "incompetent physicians" etc). There is a case for editorial (and user) tools to generate classes of entities experimentally to determine their effect on the database -- and to eliminate them if necessary. With problems it is as much the pattern created by these negative operators which structures the database as the pattern of subjects.

6. Augmenting editorial conceptual skills

This "hypertext editing" process is as much art as science and can only be developed with experience and natural aptitude. What is currently sought is a way of augmenting the conceptual skills of the editor. These can perhaps best be described through one or more metaphors:

• (a) Knitting: There is a definite sense of knitting together entities into patterns. The challenge for the editor is to hold and refine that pattern in relation to entities that might or might not be drawn into it. To be able to visualize the pattern would be of great value.
• (b) Gardening: There is a sense of planting entities over a conceptual space in an aesthetic and functional relationship to one another. Entities and networks may be pruned or given scaffolding on which to develop when new information becomes available.
• (c) Ecological succession: There is a sense in which the database is always evolving and never complete. At any time, the entities are arranged and related like plants and animals within an ecosystem. But the relationships are changing continually -"hardy" pioneer entities providing niches for more specialized and subtle entities, the entire system elaborating in diversity and complexity as time goes on. The challenge for the editor is to optimize the degree of integrity and system resilience in anticipation of unforeseeable future developments.

Clearly such hypertext editing is a new kind of skill which will be basic to the development of many future databases and their multimedia access. Many of the required features are relevant to the access needs of sophisticated end-users.

7. Editing network representations

The capacity to edit networks as graphic visual representations would offer many advantages, provided that such changes are integrated back into the database.

This facility represents a combination of hypertext editing and the "networking mapping" facilities described below. As in CAD, PCB, CASE and other design packages, the editor is dealing with relationships in a visualized form, as opposed to linear text. The editor "draws" in new relationships, moves relationships, or deletes them.

8. Use of network maps as alternative indexes

Whether as editor or end-user, related facilities would be the use of the network map as an index, permitting the user to navigate through the network. Text relating to particular nodes could be called up when required, and then edited or annotated according to access rights.