| dc.description.abstract | With the digitization of information, discoveries of events that previously took much human effort can now be found automatically. As example, we investigate several scandals in the art and antiques area that occurred between 1985 and 2005. In these events, the auction house Sotheby's was suspected to accept or even help the trading of smuggled paintings or antiques and the famous Getty Museum was exposed as purchasing antiques linked to treasure hunters. Discovering these secrets required the hard work of journalists, detectives, TV producers, and so on. The investigators were involved in illegal trades and various dangerous situations during their process of investigation. In comparison, today, with the access to digital version of large datasets, we are able to discover similar events using computationally-based techniques without the high risk and the cost of human labour needed before.
This thesis introduces our tool for extracting keywords, terms and peoples' names from news articles, books, and marking them on an interactive map. We use the New York Times as the main resource, extract location terms in each news articles using Gazetteer, extract keywords and people's names in each articles and reduce ambiguity using WordNet. Combining them, we are able to form location-keyword-time pairs for each articles, and together they form a database. Then we build an interactive map based on the database. The map is able to show the relationships between location and keywords. The linkages between two or more people or locations is able to show on the map. The demonstration was able to perform similar detection process as those journalists did in the late 90s.
The paper also introduces additional findings during the examination of the original datasets. As a news media outlet based in New York, we see evidence that the New York Times turns out to focus much more on New York City and the United States compared with other countries. With the extraction of locations inside the articles, we were able to see the distribution of articles mentioning different countries differs a lot when comparing the different continents. Our visualization also shows how locations names were changed throughout time, and how the terms people use describing a certain object changes. | en |
