A join project by the UNHCR and UN Global Pulse, exploring sentiments around forced displacement, used social media data to gather a real-time response to the ongoing situation and better understand the benefits of big data. The initial analysis aimed to learn about refugees’ behaviour towards each other and towards service providers in Europe. However, following a rise in media attention and negative sentiments in the aftermath of the terror attacks across the continent, researchers decided to shift focus and analyse the relations between forced migrants and transit and host communities. Researches hoped to use findings to improve operational services across Europe.
Social media presents a unique opportunity for analysis in situation such as the Europe Refugee Emergency where large communities move around, often untracked, preventing traditional ways of data gathering (eg. surveys). Data gathered from platforms such as Twitter or Facebook offer a real-time perspective on the events from a diverse group of users in multiple locations. The joint project used data collected from Twitter, analysing different blocks of the tweets such as hashtags or keywords to create a taxonomy. Next, the system searched for posts containing the identified words or sentences which were then manually coded by the researchers. This so called ‘training database’ was the basis on which the algorithms classified new posts, adding them to the dataset. Responses were classified in three distinct categories ranging from xenophobic to neutral and finally to irrelevant.
Researchers noted the difficulties in differentiating between accents and slangs when creating the queries for the system. Given that the data only contained excerpts, it was difficult to identify the person behind the tweet. Furthermore, there appeared to be some inconsistency around languages, for instance, the monitor in Farsi was not set up as only a small sample was obtained. Therefore, while the study provided a new insight into big data analytics in relation to humanitarian issues, the project also highlighted some important limitations and issues around such analysis. It is important to remember that while social media has great methodological advantage when combined with traditional sources of data, it can nevertheless provide a full picture in itself.
UNHCR and UN Global Pulse (2017) Social Media and Forced Displacement: Big Data Analytics and Machine Learning