Think Tank Simulator: A Virtual Conversation between the Information Seeker and Domain Experts

Giancarlo Crocetti, Pace University

Abstract

Search tools used today in the life sciences industry, mainly use a query-response paradigm, returning relevant documents based on the search intent captured by keywords specified by the user. This model works very well when the search intent is specific and the user knows exactly what to look for; however, if the search tools is used for the acquisition or identification of new or unusual knowledge this model completely breaks down due to the unavoidable fuzziness contained in the query and the consequent inability of the ranking system to identify relevant content. The rich cognitive process necessary for the acquisition of new or unusual knowledge requires a more exploratory approach and one model that has been used successfully throughout contemporary history, in many and diverse fields, is the Think Tank. In a think tank setting, experts from different fields get together and converse so to provide feedback related to the problem at hand. During such conversations, important connections start to arise and eventually new and unusual findings are identified as important contribution toward the solution of the problem. In this study we have created a tool, the Think Tank Simulator, which simulates a think tank setting where scientific papers, from the PubMed ® Central Open Access Subset, are used as proxies to experts and indexed by a search software system. The fuzzy search intent of a user is then used to retrieve as many documents as possible that are related to the problem being researched and used to simulate conversations among the experts. Through the use of biomedical ontologies and statistical tools we then try to identify unusual and previously unknown connections using terms appearing in such conversations and display them back to the user using a graph-based visual metaphor as a tool to provide a meaningful exploration of related items identified by the system. The study then surveyed users from different departments—from scientists in the Research & Development department to marketing research professionals—to gain an understanding on how they search for information and especially to measure the level of satisfaction of using traditional search tools for the identification and acquisition of new and unusual knowledge. All participants in this study have produced reports using search tools based on a query-response paradigm, and their findings were compared with the output generated by the Think Tank Simulator. The results provided by the simulator were collected during a follow-up interview with the intent to determine the user’s perception of the tool’s effectiveness in identifying new or unusual findings. The technique demonstrated in this study was found to be associated to a better outcome for the discovery of new or unusual information in scientific literature. Moreover, throughout the study we have uncovered several other findings related to the use of specific Part Of Speech (POS) entities on portion of the documents where the match occurred: lemmatized nouns were associated to better outcomes.

Subject Area

Information Technology|Information science|Computer science

Recommended Citation

Crocetti, Giancarlo, "Think Tank Simulator: A Virtual Conversation between the Information Seeker and Domain Experts" (2015). ETD Collection for Pace University. AAI3731733.
https://digitalcommons.pace.edu/dissertations/AAI3731733

COinS