What
SolrSherlock is intended to be a Watson-like framework for research, question answering, and other purposes. The project is predicated on the use of a topic map as its primary information store, its primary resource for performing its tasks.
Why
Drawing partially from issues such as infoglut, large and growing collections of not-well-organized information resources, we differentiate between reading unstructured text at question-answering time and doing so before question-answering time. Simply stated, the thesis here is that the use of a pre-compiled view of the literature will improve performance during task time in SolrSherlock.
How
Our approach is to build into SolrSherlock a variety of
machine reading agents which continuously build and maintain a topic map. An
Apache Solr or
SolrCloud installation hosts the topic map.