Narrative extraction: Analyzing the world’s narratives through natural language understanding

Description

Nobel laureate Robert Shiller is one of few economists to have predicted both the Great Recession of 2007–2009 and the dot-com crash. Throughout his career, Shiller has introduced and supported groundbreaking ideas in economics, and in 2017, he presented "Narrative Economics" at the National Bureau of Economic Research (NBER). The paper introduced two key concepts:

1. Macroeconomic fluctuations are not only influenced by economic indicators but also by the origination and spread of popular narratives.
2. Narratives spread along a pattern similar to contagious disease epidemics. Shiller proposes using epidemiological disease modeling techniques to understand the impact and dynamics of these narratives.

The term “narratives” refers to stories or ideas, particularly those of human interest and emotion that are conveyed in communication. The human brain has always been highly attuned to narratives, whether factual or not, to justify ongoing actions. Stories motivate and connect activities to deeply felt values and needs. When narratives “go viral,” they spread far and wide and are often accompanied by deep sociopolitical and economic change. The 1920–21 Depression, the Great Depression of the 1930s, the Great Recession of 2007–9, and today’s contentious political-economic situation are all considered the results of the popular narratives of their respective times.

The interpretation of narratives is a subjective human phenomenon and is difficult to study scientifically. Narrative extraction is the first step to this end. Narrative extraction is particularly relevant today, when “fake news” and the manipulation of social media propagate stories that have motivated human action, for example, during the 2016 US election or the Brexit referendum. Beyond these not-so-trivial matters, research in this area will assist corporations in understanding brand value dynamics and the virality of products and will enable improved modeling of a product’s diffusion of innovation curve. Spreading narratives have been creating cultural change long before the internet or the advent of social media. However, the dynamics of this phenomena have both accelerated and become more decentralized. Narratives traverse the “Dunbar’s number” threshold far more frequently than ever before and appear to influence a greater multiplicity of social groups.

Naveed Ghaffar and Rashed Iqbal outline a framework that uses natural language understanding to extract and analyze narratives in human communication in order to help us gain a better understanding of how “viral” narratives emerge as well as their likely impact. The approach involves:

• The automation of narrative extraction by identifying and extracting key concepts around a theme of conversation. Narratives are sourced through data repositories such as GDELT.
• Understanding by generating an ontological framework for concepts and associations. Natural language enderstanding is said to operate at multiple levels, the lowest being the syntax level followed by semantic and then pragmatic levels. Narrative extraction lies next to the pragmatic level adding human subjectivity into the mix.
• Categorizing narrative types by conveying ideas through the delivery mechanism of stories. Through their research, Naveed and Rashed have developed the key categories for these narratives types. Each category adds rich context to an idea and allows the modeling of the likely propagation effect or “stickiness” of the narrative itself.
• Searching for and identifying narratives from the tens of millions of popular narratives flowing across the world at any moment. Narrative extraction enables search of these narratives specific to a geolocation, around a concept, or for a product, as well as across the popularity of narratives or their predicted impact to influence types of human action.