Who Blames Whom in a Crisis? Detecting Blame Ties from News Articles Using Neural Networks

Abstract

Blame games tend to follow major disruptions, be they financial crises, natural disasters or terrorist attacks. To study how the blame game evolves and shapes the dominant crisis narratives is of great significance, as sense-making processes can affect regulatory outcomes, social hierarchies, and cultural norms. However, it takes tremendous time and efforts for social scientists to manually examine each relevant news article and extract the blame ties (A blames B). In this study, we define a new task, Blame Tie Extraction, and construct a new dataset related to the United States financial crisis (2007)(2008)(2009)(2010) from The New York Times, The Wall Street Journal and USA Today. We build a Bi-directional Long Short-Term Memory (BiLSTM) network for contexts where the entities appear in and it learns to automatically extract such blame ties at the document level. Leveraging the large unsupervised model such as GloVe and ELMo, our best model achieves an F1 score of 70% on the test set for blame tie extraction, making it a useful tool for social scientists to extract blame ties more efficiently.

Who Blames or Endorses Whom? Entity-to-Entity Directed Sentiment Extraction in News Text

“…We first construct previous approaches proposed for blame relationship detection. Liang et al (2019) proposed three neural models that predict the classes of blame relationships in a given text: p → q, p ← q, and no relationship. We extend their approaches to the five classes of directed sentiment in our dataset.…”

“…The dictionary-based approach filters in sentences containing positive or negative keywords from a pre-defined dictionary. Starting from the blame-related keywords (Liang et al, 2019), we extended the dictionary by adding their synonyms and antonyms. While this method is effective in sampling from an unbalanced dataset, it excludes sentences that do not explicitly mention a blame or support keyword.…”

Social Media Attributions in the Context of Water Crisis

“…Methodologically, our work is closely related with automatic extraction of blame ties (Liang et al, 2019). Similar to Liang et al (2019), we seek to extract causal ties (Miwa and Bansal, 2016) between a crisis and different possible factors. However, unlike the present work, Liang et al (2019) focused on a clean corpus obtained from three major US newspapers.…”

“…Similar to Liang et al (2019), we seek to extract causal ties (Miwa and Bansal, 2016) between a crisis and different possible factors. However, unlike the present work, Liang et al (2019) focused on a clean corpus obtained from three major US newspapers. In contrast, we embrace the challenge of detecting attribution ties from noisy, social me-dia data which involves the following challenges.…”

“…For example, both the comments, 'we must protect our forests plant more trees' and 'just rewind and see how many trees have vanished over the years to accommodate more space for buildings and malls' deemed deforestation responsible for water crisis but have different ways of expressing it. In contrast, Liang et al (2019) dealt with a set of well constructed entities that are easy to detect in a sentence due to their crisp word boundaries.…”