This paper discusses the results of investigating simple, cognitive-based approaches to search. The emphasis is placed on simplicity, and determining if a simple ranking measure is sufficient for improved search precision. The measures chosen are concept-based since concept and context-based search improves precision. These results provide direction on the need for more complicated methods. If a simple, yet effective, distance measure is found for rank-ordering search results for improved precision, then approaches may be feasible for improving search precision in a shorter period of time at less cost. Moreover, the methods investigated use a natural language interface that enables far more complicated criteria while remaining intuitive to the casual user. Furthermore, these criteria better reflect search requirements than keywords alone. Two cognitive measures were investigated: a topology-based measure, and a cogency-based measure, both using a medical ontology. The corpus for testing search precision was sampled from NLM publication abstracts, and search results were scored by a physician. Results indicate that improving search precision via the simple use of these two measures, even though related to cognition, are insufficient for significant improvements in search precision. While a simple ranking metric is preferred, the results suggest that efforts to improve search precision are better spent on more complicated methods, for example, neural network-based approaches. These results aid in guiding future research.
- Artificial intelligence,
- Natural language processing systems,
- Semantics,
- Cognitive searches,
- Health care informatics,
- Search,
- Search relevancy,
- Semantic search,
- Ontology,
- Healthcare informatics
Available at: http://works.bepress.com/donald-wunsch/63/