Projects

SKR research efforts support both SemRep development and the use of semantic predications in innovative biomedical information management applications, particularly for literature-based discovery and hypothesis generation.

SemRep development

SemRep was originally devised for the clinical domain and was subsequently extended to genetic etiology and pharmacogenomics. We have recently devised a domain-independent methodology that allows us to leverage existing UMLS knowledge by adapting well-known ontology engineering phases and seamlessly integrating them with the knowledge sources afforded by the UMLS, extending coverage within a newly defined semantic space [1]. The ontological and terminological extensions implemented in the system to extend SemRep to other domains have been successfully deployed in medical informatics knowledge processing, disaster information management, and public health promotion. Semantic coverage in SemRep is currently limited to what is asserted in the texts. Authors, however, also express subjectivity, such as beliefs, speculations, opinions, intentions, and desires. Furthermore, they link statements of various kinds to form a coherent discourse. Recent research in the SKR project underpins a major expansion of SemRep in order to interpret this extra-propositional meaning [2]. Such processing will enhance the value of semantic predications by determining whether they are supported by strong, compelling evidence, based on their factual status and explicitness of evidence. This extension of semantic coverage will strengthen the use of SemRep predications for both literature-based discovery and hypothesis generation.

SemRep applications

To support more effective biomedical information management, we have developed Semantic MEDLINE (SemMed), which integrates document retrieval, advanced natural language processing, automatic summarization, and visualization into a single Web portal [3]. Ongoing research aims at developing and integrating additional functions for this application.

SKR research exploits predications and graph theory for automatic summarization of biomedical text. One project used degree centrality to measure connectedness in a graph, for summaries based on more than 500 citations [4]. In another project, we developed a clique-clustering method to automatically summarize graphs of predications produced from large numbers of PubMed titles and abstracts (more than 10,000 citations) [5]. Themes contained in the summary were identified with a hierarchical clustering algorithm based on common arguments shared among the cliques.

Significant research has been devoted to developing and applying the literature-based discovery paradigm using semantic predications. We focused on the development of "discovery browsing," a method that guides the user through the research literature on a specified phenomenon [6]. Poorly understood relationships may be explored through novel points of view, and potentially interesting relationships need not be known ahead of time. In a process of "cooperative reciprocity" the user iteratively focuses on the output generated by the system, thus controlling the large number of relationships often produced in literature-based discovery systems.

References

  1. Rosemblat G, Shin D, Kilicoglu H, Sneiderman C, Rindflesch TC. (2013). A methodology for extending domain coverage in SemRep. Journal of Biomedical Informatics. 2013 Dec;46(6):1099-107.
  2. Kilicoglu H, Rosemblat G, Rindflesch TC. (2017). Assigning factuality values to semantic relations extracted from Biomedical Research Literature. PLoS One. 2017; 12(e): e0179926.
  3. Rindflesch TC, Kilicoglu H, Fiszman M, Rosemblat G, Shin D. (2008). Semantic MEDLINE: An advanced information management application for biomedicine. Information Services & Use, 31, 15-21.
  4. Zhang H, Fiszman M, Shin D, Miller CM, Rosemblat G, Rindflesch TC. (2011). Degree centrality for semantic abstraction summarization of therapeutic studies. Journal of Biomedical Informatics 2011 Oct; 44(5):830-38.
  5. Zhang H, Fiszman M, Shin D, Wilkowski B, Rindflesch TC. (2013). Clustering cliques for graph-based summarization of biomedical research literature. BMC Bioinformatics 14:182.
  6. Cairelli MJ, Miller CM, Fiszman M, Workman TE, Rindflesch TC. (2013). Semantic MEDLINE for discovery browsing: Using Semantic predications and the literature-based discovery paradigm to elucidate a mechanism for the obesity paradox. Proceedings AMIA Annual Symposium 2013;164-73.