Predicting the Rise and Fall of Scientific Topics from Trends in their Rhetorical Framing

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 1170–1180, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics Predicting the Rise and Fall of Scientific Topics from Trends in their Rhetorical Framing Vinodkumar Prabhakaran Stanford University vinod@cs.stanford.edu William L. Hamilton Stanford University wleif@stanford.edu Dan McFarland Stanford University dmcfarla@stanford.edu Dan Jurafsky Stanford University jurafsky@stanford.edu Abstract Computationally modeling the evolution of science by tracking how scientific top- ics rise and fall over time has important implications for research funding and pub- lic policy. However, little is known about the mechanisms underlying topic growth and decline. We investigate the role of rhetorical framing: whether the rhetori- cal role or function that authors ascribe to topics (as methods, as goals, as results, etc.) relates to the historical trajectory of the topics. We train topic models and a rhetorical function classifier to map topic models onto their rhetorical roles in 2.4 million abstracts from the Web of Science from 1991-2010. We find that a topic’s rhetorical function is highly predictive of its eventual growth or decline. For exam- ple, topics that are rhetorically described as results tend to be in decline, while top- ics that function as methods tend to be in early phases of growth. 1 Introduction One of the most compelling research questions in the computational analysis of scientific literature is whether the vast collections of scientific text hold important clues about the dynamics involved in the evolution of science; clues that may help predict the rise and fall of scientific ideas, meth- ods and even fields. Being able to predict sci- entific trends in advance could potentially revolu- tionize the way science is done, for instance, by enabling funding agencies to optimize allocation of resources towards promising research areas. Figure 1: Example abstract snippet. The abstract rhetorically frames the stem cells topic as the OBJECTIVE of the research, while the animal models topic functions as the research METHOD. Prior studies have often tracked scientific trends by applying topic modeling (Blei et al., 2003) based techniques to large corpora of scientific texts (Griffiths and Steyvers, 2004; Blei and Laf- ferty, 2006; Hall et al., 2008). They capture sci- entific ideas, methods, and fields in terms of top- ics, modeled as distributions over collection of words. These approaches usually adopt a de- contextualized view of text and its usage, associ- ating topics to documents based solely on word occurrences, disregarding where or how the words were employed. In reality, however, scientific ab- stracts often follow narrative structures (Crookes, 1986; Latour, 1987) that signal the specific rhetor- ical roles that different topics play within the re- search (Figure 1). The rhetorical role of a topic is the purpose or role it plays in the paper: as its background (scientific context), its objective/goal, the data employed, the design or method used (mode of inference), the results (what is found) or the conclusions (what they mean). 1170 RATIONALE: Neonatal ibotenic acid lesion of the ventral hippocampus was proposed as a relevant …

Content truncated to 500 words