DOE rolls out semantic search for science research
It's an axiom of research that a database is only as good as the search tool being used to explore it. The Energy Department’s Office of Scientific and Technical Information, therefore, has taken a step toward better databases in releasing SciTech Connect, a reference tool that helps users browse DOE's core reference collections using semantic search techniques.
Semantic search is the Holy Grail of computer-assisted research, intended to improve the quality of search results. Instead of using ranking algorithms as in a Google search, semantic search considers the context of a query, including word variation and natural language questions, to provide more relevant search results.
In unveiling the tool, OSTI director Walter Warnick said SciTEch also achieves the agency's goal of re-circulating research findings between DOE and the scientific community.
"With SciTech Connect," he said, "we are [making] DOE R&D results easier to retrieve and thereby [able to] better serve our dual core mission – getting DOE results out to the scientific community and beyond, and getting the community’s results into DOE.”
The SciTech tool contains full text documents of two earlier DOE scientific reference resources, the Information Bridge and the agency's Energy Citations database. Together, the new reference system comprises 65 years of energy-related scientific citations, DOE said, including 2.5 million citations and 1.4 million journal articles.
SciTech Connect will eventually replace the Information Bridge and Energy Citations databases in a transition period that should be completed by July.
DOE said SciTEch uses a semantic search approach called keyword-to-concept mapping. When a keyword-based query is made, the system maps the query to other associated terms, to produce a taxonomy of narrow and related concepts.
The technique enables the search engine to "make use of the logical relations among concepts in different scientific documents, regardless of whether those documents use standard [terms] to express those concepts," DOE said.
It is an "especially effective search approach when a person is truly is researching a topic rather than trying to navigate to a particular destination.
SciTech Connect has a other features, including in-document search; word clouds; and tools for personalization, allowing users to save searches, define alerts and create and manage document libraries.
Posted by Paul McCloskey on Apr 19, 2013 at 9:39 AM