Exploring the Usefulness of DatAR for Finding Fruitful Experiments in Neuroscience Literature

Abstract

Neuroscientists need to analyze a vast amount of literature to find potentially fruitful experiments. Topic-based literature exploration is a useful means to analyze many publications simultaneously because it provides an overview of the relations between topics: the co-occurrence of brain regions and brain diseases topics within the same sentences of a publication often implies a relation between them. In order to incorporate DatAR into the daily workflow of neuroscience literature exploration, multiple functionalities should be provided to complete tasks with full scope. Neuroscientists have indicated that filtering topics is useful, involving at least two rounds and multitasking, including identifying, comparing, and verifying identified relations. Therefore, our goal is to investigate the extent to which multiple functionalities are useful for filtering topics.

We follow a user-centered design approach. We (i) identify the user tasks of filtering topics is indeed useful with neuroscientists; (ii) identify the representative tasks for filtering topics; (iii) design the evaluation approach. The evaluation is first conducted using the initial version of DatAR (V1), followed by the iteration of multiple functionalities based on the user requirements collected from the first round of evaluation. The developed compare version of DatAR (V2) introduces five additional aspects, which are intended to improve the accuracy and sufficiency of the identified relations provided by multiple functionalities. After conducting the second round of evaluation based on V2 with six neuroscientists, the results indicate valuable insights addressing our original goal of investigating the usefulness of multiple functionalities for filtering topics. Specifically, the findings show that (i) the identified relations provided by multiple functionalities are useful for filtering topics, directly supporting our goal; (ii) the usefulness of multiple functionalities could be enhanced by providing more sufficient, latest, and closely related topics, further aligning with the goal of maximizing filtering effectiveness; and (iii) enabling the recording and comparison of differences in identified relations across multiple rounds of topic filtering, improving the visualization of identified relations in complex multitasking scenarios, and offering guidance with sufficient detailed information for filtered topics could support the integration of multiple functionalities into the daily workflow of neuroscience research. These results collectively validate and extend our goal by pinpointing specific improvements that could facilitate effective topic filtering.