Introduction
In the late 1980s, as libraries transitioned toward electronic storage, a significant bottleneck emerged: "indeterminism." This term describes the gap between how users describe their information needs and how documents are actually indexed. This research identifies three primary causes for this gap: source ignorance (not knowing which databases exist), classification gaps (misunderstanding indexing schemes), and query ambiguity (fuzzy problem definitions). While human reference librarians bridge these gaps through interactive dialogue, early automated systems lacked this "cognitive touch." This paper formalizes the librarian's mental process into a computational model to serve as a blueprint for intelligent research assistants.
Methodology: The Two-Phase Process Model
Through 30 qualitative field observations of interactions between librarians and users, the researchers synthesized a two-phase process model:
- Handle Search: The librarian acts as a "traffic controller." Instead of searching for specific documents immediately, they first identify the appropriate "handle"—a specific index or database. This phase involves "slot-filling" (extracting attributes like geographical area and time frame) and "stereotyping" (classifying the user based on education level and query scope).
- Document Search: Once a handle is selected, the librarian assists the user in translating natural language into "official" indexing terms using thesauri and Boolean operators.
Key Mechanisms: Slot-Filling and Stereotyping
The model's efficiency relies on reducing the search space before document retrieval begins. "Slot-filling" extracts four core attributes: area covered, currency, type of information, and field of applicability.
Simultaneously, the "stereotyping" mechanism maps a user's education level (from Freshman to Ph.D.) against their query scope (from class papers to theses). This allows the system to prioritize information types—for example, directing a Ph.D. student toward journal articles while steering a freshman toward general textbooks. This approach follows the "Minimum Superset Heuristic," aiming to provide sufficient information without overwhelming the user with irrelevant data.
Results and Discussion
The formalized model demonstrates that "goal-directed questioning" can significantly reduce search uncertainty. By identifying the four core attributes of a "handle" and applying heuristics like the "Partial Match Heuristic" (ranking sources when no perfect match exists), the model simulates the expert librarian's ability to turn a "fuzzy" query into a precise search strategy. The research highlights that uncertainty is two-sided: indexing uncertainty (how documents are labeled) and searching uncertainty (how needs are described). The librarian—and this model—acts as the essential "term translator" between these two worlds.
Limitations and Future Work
The authors acknowledge that this model is a blueprint for standard consultations and may not replace human experts in highly sophisticated or unusual cases. The study's scope was limited to online access tools, excluding physical media. The next phase of research involves the actual implementation of this "intelligent assistant" within university library systems to empirically test its ability to reduce the burden on human staff.
Summary updated