Multi-Label and Multilingual News Framing Analysis

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8614–8624 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 8614 Multi-Label and Multilingual News Framing Analysis Afra Feyza Akyürek1, Lei Guo2,1, Randa Elanwar3, Prakash Ishwar4,1, Margrit Betke1 and Derry T. Wijaya1 1Department of Computer Science, Boston University 2College of Communication, Boston University 3Electronics Research Institute, Egypt 4Department of Electrical and Computer Engineering, Boston University {akyurek,guolei,relanwar,pi,betke,wijaya}@bu.edu Abstract News framing refers to the practice in which aspects of specific issues are highlighted in the news to promote a particular interpretation. In NLP, although recent works have studied fram- ing in English news, few have studied how the analysis can be extended to other languages and in a multi-label setting. In this work, we explore multilingual transfer learning to de- tect multiple frames from just the news head- line in a genuinely low-resource context where there are few/no frame annotations in the tar- get language. We propose a novel method that can leverage elementary resources consisting of a dictionary and few annotations to detect frames in the target language. Our method per- forms comparably or better than translating the entire target language headline to the source language for which we have annotated data. This work opens up an exciting new capabil- ity of scaling up frame analysis to many lan- guages, even those without existing translation technologies. Lastly, we apply our method to detect frames on the issue of U.S. gun vio- lence in multiple languages and obtain excit- ing insights on the relationship between differ- ent frames of the same problem across differ- ent countries with different languages. 1 Introduction The worldwide image of the United States has dropped precipitously during the past few years (Wike et al., 2018). Among other factors, the in- creasing number of gun violence incidents appears to affect the U.S. reputation abroad. Whenever a fatal mass shooting happens, it often attracts sig- nificant international news attention. While the domestic U.S. news media often links gun violence to individual shooters’ mental illness (DeFoster and Swalve, 2018; Liu et al., 2019), foreign media may attribute it to U.S. gun policy and its gun culture e.g., (Atkinson, 2019). This phenomenon is known as media framing, which is the process of selecting “some aspects of a perceived reality and [making] them more salient in a communicating text, in such a way as to promote a particular problem defini- tion, causal interpretation, moral evaluation, and/or treatment recommendation for the item” (Entman, 1993). When foreign media frame the gun vio- lence issue in a way to depict the U.S. as an unsafe and undesired place, it erodes the country’s “soft power” (Nye Jr, 2004). Evaluating how different countries frame the U.S. gun violence issue will enrich our understanding of the U.S. soft power in particular and international relations in general. In this work, we develop a multilingual approach to automatically detect frames in news coverage of different languages, thus facilitating the analysis of how different countries with different …

Content truncated to 500 words