Effective distributed representations for academic expert search

Abstract

Expert search aims to find and rank experts based on a user's query. In academia, retrieving experts is an efficient way to navigate through a large amount of academic knowledge. Here, we study how different distributed representations of academic papers (i.e. embeddings) impact academic expert retrieval. We use the Microsoft Academic Graph dataset and experiment with different configurations of a document-centric voting model for retrieval. In particular, we explore the impact of the use of contextualized embeddings on search performance. We also present results for paper embeddings that incorporate citation information through retrofitting. Additionally, experiments are conducted using different techniques for assigning author weights based on author order. We observe that using contextual embeddings produced by a transformer model trained for sentence similarity tasks produces the most effective paper representations for document-centric expert retrieval. However, retrofitting the paper embeddings and using elaborate author contribution weighting strategies did not improve retrieval performance.

A New Neural Search and Insights Platform for Navigating and Organizing AI Research

“…We then derive the experts from the sets of authors of these papers using an approach where each retrieved paper contributes an exponentially weighted vote for an author, with a factor that reduces the bias towards highly prolific authors. Our experiments, described in detail in (Berger et al, 2020) show that these modern Transformer-based contextualized embeddings outperform TF-IDF and LSI-based document representations on this task.…”

An Explanatory Query-Based Framework for Exploring Academic Expertise

“…In addition to a cross-validation evaluation, the authors also compared the extracted keyphrases of papers with the ones supplied by the papers' authors. The focus of [Berger et al, 2020] was on analysing Computer Science papers for academic expert retrieval. Their method relies on Sentence-BERT [Reimers and Gurevych, 2019] which is used on the concatenation of the title and the abstract to represent the paper.…”

Multi-Vector Models with Textual Guidance for Fine-Grained Scientific Document Similarity

“…The ability to identify similarity across documents in large scientific corpora is fundamental for many applications, including recommendation (Bhagavatula et al, 2018), exploratory or analogical search (Hope et al, 2017(Hope et al, , 2021bLissandrini et al, 2019), paper-reviewer matching (Mimno and McCallum, 2007;Berger et al, 2020) and many more uses.…”