Using Hierarchical Task Network Planning Techniques to Create Custom Web Search Services over Multiple Biomedical Databases

Abstract

We present a novel method to create complex search services over public online biomedical databases using hierarchical task network planning techniques. In the proposed approach, user queries are regarded as planning tasks (goals), while basic query services provided by the databases correspond to planning operators (POs). Each individual source is then mapped to a set of POs that can be used to process primitive (simple) queries. Advanced search services can be created by defining decomposition methods (DMs). The latter can be regarded as “recipes” that describe how to decompose non-primitive (complex) queries into sets of simpler subqueries following a divide-and-conquer strategy. Query processing proceeds by recursively decomposing non-primitive queries into smaller queries, until primitive queries are reached that can be processed using planning operators. Custom web search services can be created from the generated planners to provide biomedical researchers with valuable tools to process frequent complex queries.