We present the IUCL system, based on supervised learning, for the shared task on stance detection. Our official submission, the random forest model, reaches a score of 63.60, and is ranked 6th out of 19 teams. We also use gradient boosting decision trees and SVM and merge all classifiers into an ensemble method. Our analysis shows that random forest is good at retrieving minority classes and gradient boosting majority classes. The strengths of different classifiers wrt. precision and recall complement each other in the ensemble.

Calls for interdisciplinary collaboration have become increasingly common in the face of large-scale complex problems (including climate change, economic inequality, and education, among others); however, outcomes of such collaborations have been mixed, due, among other things, to the so-called "translation problem" in interdisciplinary research. This article presents a potential solution: an empirical approach to quantitatively measure both the degree and nature of differences among disciplinary tongues through the social and epistemic terms used (a research area we refer to as discourse epistemetrics), in a case study comparing dissertations in philosophy, psychology, and physics. Using a support-vector model of machine learning to classify disciplines based on relative frequencies of social and epistemic terms, we were able to markedly improve accuracy over a random selection baseline (distinguishing between disciplines with as high as 90% accuracy) as well as acquire sets of most indicative terms for each discipline by their relative presence or absence. These lists were then considered in light of findings of sociological and epistemological studies of disciplines and found to validate the approach's measure of social and epistemic disciplinary identities and contrasts. Based on the findings of our study, we conclude by considering the beneficiaries of research in this area, including bibliometricians, students, and science policy makers, among others, as well as laying out a research program that expands the number of disciplines, considers shifts in socio-epistemic identities over time and applies these methods to nonacademic epistemological communities (e.g., political groups).

This study explores how male and female users of Voicethread.com, an interactive multimodal web 2.0 platform that allows asynchronous commenting via text, audio, and video, communicate and perform identity through self-expression in different semiotic modes. A quantitative computer-mediated discourse analysis of three public English-language Voicethreads found that in video and audio comments, both genders express more positive attitudes; they are also more self-conscious and ego-focused. The text comments express more neutral and negative attitudes, especially when written by males, but they are also more socially interactive. With few exceptions, female communication patterns resemble those for audio and video, while male communication patterns resemble those for text. We propose explanations for these findings and discuss their implications for identity performances in interactive multimodal environments.

No abstract