Text mining is a technique used to rapidly examine large corpora of texts, such as online news reports or other web content, to identify useful nuggets of information. The emergence of this field within the artificial intelligence (AI) domain has already been applied successfully in various contexts such as business intelligence, opinion mining and sentiment analysis. Such technology can also be used to analyse natural disasters and their impact on society.
Research we have undertaken exploited such text mining tools to look for earthquake-related news. The system runs automatically, mapping the news sources and earthquake locations in near real-time, and verifies the information with an international seismological centre to confirm the news report.
One of the novel approaches adopted is that the software can filter through multi-lingual websites such as Italian, Mexican, Chilean and Chinese texts. This gives the added advantage of reducing news bias originating from English written newspapers only.
The developed prototype, named QuakeNews Analyser, includes components used to detect and track events, filter content, cluster news articles, extract information, cross-validate and visualise the mined data.
It paves the way for a better understanding of the coverage given by different news agencies worldwide with respect to earthquake events. This includes how quickly news agencies react to earthquake events, how accurate news agencies are when reporting, how long an earthquake event remains mentioned in successive news items and how far away from the earthquake location it is reported.
More importantly, this research managed to successfully extract different forms of information, including ‘cardinal numbers’ such as the magnitude of the earthquake, the number of casualties, quantifiable structural damage caused, and event dates (in different formats) and geographical locations. Generally, this type of data is challenging to extract because it is often expressed in different ways in unstructured text. The results are then visualised on a user-friendly dashboard showing the geographical distribution of news articles for earthquake events, as well as a search facility.
Through this work, different types of analysis can be carried out. For instance, one can note that the Papua New Guinea 7.5-magnitude earthquake on February 25, 2018, was the most mentioned event (50 articles over a span of 48 days). On the other hand, Taiwan’s 6.4-magnitude seismic event on February 6, 2018, was the most reported event within a short timeframe (47 articles over a period of five days). Interestingly, The Guardian was found to be the least accurate when reporting the magnitude of the event, while the Shanghai Daily was found the most accurate when reporting the magnitude value.
Stephen Camilleri is a recent Master’s graduate in Artificial Intelligence (Big Data) from the Faculty of Information & Communication Technology, University of Malta. He was supervised by Dr Joel Azzopardi and Dr Matthew Agius, and recently won an outstanding student poster award for this work during this year’s European Geosciences Union (EGU) General Assembly for his application in the context of geosciences (https://bit.ly/2qkSgzD). This work has also been published in the IEEE Conference on Artificial Intelligence and Knowledge Engineering (IEEE AIKE 2019) - https://ieeexplore.ieee.org/document/8791724.
Did you know?
• The marathon started in 1896 in Athens, Greece, a race from Marathon – northeast of Athens – to the Olympic Stadium, a distance of 42.195 kilometres.
• The first women’s marathon in the Olympics was held in 1984.
• The first New York City Marathon was held in 1970 and cost $1 for the 127 participants.
• Whilst Marie-Louise Ledru was the first woman to be credited with completing a marathon in 1918, it was Violet Piercy who was first officially timed, in 1926.
For more trivia see: www.um.edu.mt/think
• More than 11,000 scientists from around the world have just signed a declaration whereby in clear and unequivocal terms, they state that the planet Earth is facing a climate emergency and that scientists have a moral obligation to warn humanity of any catastrophic threat and to ‘tell it like it is’.
• Scientists have discovered how a potent bacterial toxin is able to target and kill MRSA, a well-known superbug commonly found in hospitals. The toxin, an enzyme called Lysostaphin, specifically recognises MRSA cell walls and quickly causes its rapid breakdown.
For more soundbites listen to Radio Mocha on Mondays at 7pm on Radju Malta and Thursdays at 4pm on Radju Malta 2 https://www.fb.com/RadioMochaMalta/