A Utility Model of Authors in the Scientific Community

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1510–1519, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. A Utility Model of Authors in the Scientific Community Yanchuan Sim School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213, USA ysim@cs.cmu.edu Bryan R. Routledge Tepper School of Business Carnegie Mellon University Pittsburgh, PA 15213, USA routledge@cmu.edu Noah A. Smith Computer Science & Engineering University of Washington Seattle, WA 98195, USA nasmith@cs.washington.edu Abstract Authoring a scientific paper is a complex process involving many decisions. We in- troduce a probabilistic model of some of the important aspects of that process: that authors have individual preferences, that writing a paper requires trading off among the preferences of authors as well as ex- trinsic rewards in the form of commu- nity response to their papers, that prefer- ences (of individuals and the community) and tradeoffs vary over time. Variants of our model lead to improved predictive ac- curacy of citations given texts and texts given authors. Further, our model’s pos- terior suggests an interesting relationship between seniority and author choices. 1 Introduction Why do we write? As researchers, we write pa- pers to report new scientific findings, but this is not the whole story. Authoring a paper involves a huge amount of decision-making that may be influenced by factors such as institutional incen- tives, attention-seeking, and pleasure derived from research on topics that excite us. We propose that text collections and associated metadata can be analyzed to reveal optimizing be- havior by authors. Specifically, we consider the ACL Anthology Network Corpus (Radev et al., 2013), along with author and citation metadata. Our main contribution is a method that infers two kinds of quantities about an author: her associ- ations with interpretable research topics, which might correspond to relative expertise or merely to preferences among topics to write about; and a tradeoff coefficient that estimates the extent to which she writes papers that will be cited versus papers close to her preferences. The method is based on a probabilistic model that incorporates assumptions about how authors decide what to write, how joint decisions work when papers are coauthored, and how individual and community preferences shift over time. Cen- tral to our model is a low-dimensional topic rep- resentation shared by authors (in defining prefer- ences), papers (i.e., what they are “about”), and the community as a whole (in responding with ci- tations). This method can be used to make predic- tions; empirically, we find that: 1. topics discovered by generative models out- perform a strong text regression baseline (Yo- gatama et al., 2011) for citation count predic- tion; 2. such models do better at that task without mod- eling author utility as we propose; and 3. the author utility model leads to better pre- dictive accuracy when answering the question, “given a set of authors, what are they likely to write?” This method can also be used for exploration and to generate hypotheses. We provide an in- triguing example …

Content truncated to 500 words