HASVRec: A modularized Hierarchical Attention-based Scholarly Venue Recommender system

Abstract

Manually selecting appropriate scholarly venues is becoming a tedious and time-consuming task for researchers due to many reasons that include relevance, scientific impact, and research visibility. Sometimes, high-quality papers get rejected due to mismatch between the area of the paper and the scope of the journal. Recommending appropriate academic venues can, therefore, enable researchers to identify and take part in relevant conferences and publish in journals that matter the most. A researcher may certainly know of a few leading venues for her specific field of interest. However, a venue recommendation system becomes particularly helpful when exploring a new domain or when more options are needed. Due to high dimensionality and sparsity of text data, and complex semantics of the natural language, journal identification presents difficult challenges. We propose a novel and unified architecture that contains a Bi-directional LSTM (Bi-LSTM) and a Hierarchical Attention Network (HAN) to address the above problems. We call the proposed architecture modularized Hierarchical Attention-based Scholarly Venue Recommender system (HASVRec), which only requires the abstract, title, keywords, field of study, and author of a new paper along with its past publication record to recommend scholarly venues. Experiments on the DBLP-Citation-Network V11 dataset exhibit that our proposed approach outperforms several state-of-the-art methods in terms of accuracy, F1, nDCG, MRR, average venue quality, and stability.