Skip to main content
Linking NeuroBridge and NeuroQuery with Deep Semantic Matching
Howard Lander, Arcot Rajasekar, Yue Wang, Matt Watson, Satya Sahoo, Jessica Turner, Jean-Baptiste Poline, Lei Wang
Presenting author:
Howard Lander
Background. The NeuroBridge (NB) project was funded by NIH to increase the findability and reuse of valuable neuroimaging data sets. The NB project uses a domain specific ontology to expand the search space beyond lexical analyses and to mediate across multiple data resources. The NeuroQuery (NQ) project is designed to produce fMRI activation maps by querying the neuroscience literature. Here we present preliminary work linking user queries at NB to relevant fMRI literature stored on NQ.

Method. Users query for data in the NB portal by selecting terms from the NeuroBridge ontology. The NQ project has its own list of terms. To integrate the projects, we created an NQ server and a Semantic Server (SS). The NQ server wraps the existing NQ search code so that it can be called by the NB portal. The SS uses Elasticsearch and SapBERT to semantically match terms in the NB ontology to NQ terms to improve the efficiency of the NQ search.

Results. The outcomes of the Semantic Server are presented to the user to select which words of the NB term they think are most likely to produce an interesting set of results from the NQ server. The results from the subsequent NQ server search are presented to the user in the NB portal.

Conclusions. Integrating various neuroimaging dataset repositories into one portal is an effective strategy for enabling the reuse of these datasets. Using semantic analyses to improve the mapping between the term sets of the repositories enhances the user experience.