This group discussed issues related to a functional model of the process of using a Knowledge Organization Systems (KOS) over a network. While the group acknowledged that there are different KOSs and that the functions might differ, discussion focused on thesauri as the most complex. Different types or phases of thesaurus use were discussed.

 1. Consultation

2. Description (Indexing)

3. Query (Search)

4. Other types of uses

The consultation phase specifically was discussed in more detail. The interactions here are complex and need to be mapped out in some detail. For example, the user first provides an entry term and/or a context in which the process of navigation should begin. This entry term may be a full concept term, a truncated concept term, or a keyword taken from a concept term; the responses that the user might want to receive vary depending on the type of entry term provided. Responses also may vary in terms of whether the entry term has a single match in the reference tool, multiple matches in the reference tool, or no match at all. The response also varies depending on whether the term is an entry term or a preferred term. In some cases, for example, an alphabetical list of thesaurus terms (with perhaps some limited amount of context such as an indicator of facet) may be the best response. In other cases, the user might want to see a fuller display of information on a single term.

What a display of information on a single term might mean was also briefly discussed. A model would need to have some way for the client to specify what "components" (relationships, notes, definitions, etc.) of information about the term were wanted, as well as how many levels of components or what combination of levels and hierarchy were wanted. The type of relationship between terms, as well as the "distance" (the degree of expansion or explosion) should be able to be specified. It was also noted that the concept of "distance" is complicated by the different levels of granularity present in different parts of a thesaurus. In addition, the differences in how users approach information retrieval indicate that different views of the thesaurus may be needed by different users for the same function.

There was also some discussion of what a functional model in fact should include. While the model needs to be driven by user needs, a standard should not specify a specific user interface (the interaction between the user and his or her own client software), but rather the interaction over the network between the software on the users' computer (the client) and the software on the computer that manages the reference tool ("the thesaurus server"). In discussing individual issues, it is not always easy to draw the line between whether a feature belongs properly to the user interface or the client-server interaction.

While one interaction takes place between the client and the thesaurus server (consultation), the other phases (query, description) take place between the client and a database server. However these interactions are not entirely independent. For example, if the database server knows about the reference tool (e.g. a bibliographic database that uses a specific classification scheme) then the client software may be able to rely on the database server to make use of the classification in doing a search, to the level of specifying the field to be searched. However if the client is accessing a thesaurus to improve searching on a database that does not use or know about the

classification scheme, then the client may need to store information from the thesaurus server, and provide the database server with a more complex search.