Backcasting Semantic Risk Indicators in Historical Travel Advisories: Enriching a Longitudinal Web Corpus

<p>Travel warnings can affect the political and economic situation in the target countries. However, data to systematically investigate these links are lacking. Previous work has introduced and validated a crawler pipeline to obtain time series of government-issued travel advisories via the Internet Archive's Wayback Machine. Governments may change the type and amount of information they provide in their advisories. For example, the United States began to provide numerical travel warning levels for individual countries and tags to categorize advisories in 2018. We illustrate an approach to using these \textit{labeled} data points to extend the variable for travel advisories before 2018 and for other issuing countries. In addition, we show how to extract additional information, such as whether the advisory concerns an entire country or specific regions or cities. This paper documents a further step towards a harmonized and unified dataset and a multilingual text corpus of government-issued travel advisories for the social sciences, humanities, and beyond.</p>