Towards Explainable Scientific Venue Recommendations

An essential part of the scientific research process is the publication of the obtained results at a suitable venue, i.e., a particular conference, workshop, or journal. The related selection problem for the best fitting scientific venue has many different aspects, such as the fit of the research topics, the prospects of acceptance, and the prestige of the venue. The complexity of the selection is further exacerbated by the growing number of publication venues, the increasing granularity of research topics, and the exponentially surging number of publications. To support researchers with this task, different methods have been proposed, e.g., based on Latent Dirichlet Allocation [9], hybrid approaches incorporating social networks [14, 13], or procedures that draw from background ontologies [20, 16]. Moreover, recent approaches based on deep learning methods achieved high accuracy in recommendations [6]. All these methods have in common that their recommendations are insufficiently explained. For example, Kobs et al. [6] solely highlight words from the input article that were essential for a recommendation. With the present work we show a new approach for recommending venues that improves on explainability. From the information a scientist provides, such as paper title, abstract and, possibly, a list of keywords, our method creates a ranking over k thematically fitting venues.