Over the past few weeks, we learned from the world press that scientists from “Microsoft” in the USA and the Technion in Haifa are working on the development of computer software that could assist in the forecasting of natural disasters, growing violence, epidemics, social unrest and other events that result in massive casualties. Closer examination of the software they developed showed that it is possible to identify signs that can detect disasters from as close as 70 to 90 percent of the time. We are not talking about a new technique, but about the new research by Kira Radinsky from the Technion and Eric Horvitz, distinguished scientist and co-director of the Microsoft Research Group laboratory at Redmond, that for the first time shows how many different sources of information can be used to improve systems and make it into something more flexible.
The software was capable of detecting the warning of a cholera outbreak in Angola through the means of information analysis on the drought that befell the country in 2006 and the storms that hit Africa in the beginning of 2007. This prediction was based on the assessment of previously compiled research, which documented cholera outbreaks following natural disasters that had previously struck areas in Angola and in other parts of the world.
Kira Radinsky researched her doctoral thesis at the Faculty of Computer Science at the Technion with Professor Shaul Markovitch as her advisor and mentor. Radinsky’s story is that of excellence and high achievements through and through.
At the age of five she began taking Karate lessons, and at age 15 – she was at the Technion studying computer science on a full scholarship (enrolled in the Technion’s Excellence Program). Upon her completion of her mandatory service in the IDF's Intelligence Corps, she successfully launched an internet-based business, worked for a start-up known as “WEBSHAKES,” and went on to work for the world renowned “Microsoft.” During these years she managed to complete her undergraduate degree, began her masters studies, and was the recipient of a number of academic prizes, among them the Technion Technological Innovation Award (from the Faculty of Computer Science), the Technion Prize for Best Academic Work (from the university), the Recipient of the Google Anita Borg Scholarship (for 20 Leading women in Computer Science), and others.
Soon after, Radinsky went on to complete her doctoral thesis at the Faculty of Computer Science, under the mentoring of Professor Shaul Markovitch. Her life at the Technion were overwhelming – in addition to her studies, Radinsky also served as a Faculty tutor (at which she also excelled at), worked for “Microsoft” and was active in the Technion Karate Team (she has a black belt). “I believe that there is a fundamental connection between karate and research,” she said. “In both instances, you must find courage and conviction to get up and resume your efforts – even if you don’t know what the outcome will bring. When you believe in yourself, you never need to give up.”
Within the framework of her master’s degree, Radinsky dealt with subject matters that had a correlation with temporal learning and artificial intelligence. In the beginning, she developed a methodology about making predictions about future events based on Internet quarries. She searched for “peaks” in search hits with regards to specific terms and expressions relating to these quarries, from which she then analyzed causal patterns surrounding different events. On the basis on these connections that she sought, she managed to successfully predict future query distribution. For example, her software was able to predict the flood of queries regarding Apple shares upon the announcement of Apple’s advertising campaign about its latest product. “In this case, we have a causal connection, but the fact that we’re dealing with a statistical mapping lets us raise unexpected connections as well,” she added. “For instance, we successfully predicted that the popularity of the word “oil” and “shares” after a day of “peak” hits on questions including the phrase a decline in dollar value.” Another interesting success, based on the same algorithms, was the prediction of hurricane storms based on the boost of quarries on the web.
From prediction based on quarry popularity, Radinsky moved on to predict future events based on computerized analysis of news archives. “We analyzed news reports published in the New York Times from the 1800’s until today, in an attempt to find causal connections. The algorithms that we developed allowed us to identify causal patterns from sequential events in the past, and predict future events based on similar incidents. As a result, we were able to foresee an increase in oil prices in the wake of a hurricane, for example.”
The technology in question was also applied to encyclopedia databases, which function differently from news archives. Here too, a significant learning curve was achieved in the identification of causal patterns – a detection that increases the projecting tools available to the user. The algorithms developed by Radinsky enables the computer also to find causal words (like “caused by,” “contributed to,” and the like) that can be used in future identification of causal relationships.
“The computer is not more intelligent that a human, but it can handle a much greater amount of factual information and find unpredictable or random connections. Since this is a learning curve (based on generalizations), we are able to find causal connections that we otherwise would not have identified through human intuition alone,” concludes Radinsky.
Recently, Radinsky moved to Zikhron Ya’akov – located in the middle of the way between her studies at the Technion, in Haifa, and her workplace, at Microsoft, Herzliya. As part of her doctoral thesis, she has continued her involvement in this area of future event prediction, based on semantic analysis and temporal learning, and hasn’t ceased her practice of karate and other hobbies, which include tennis, reading and writing poetry, salsa dancing, running and physical training at her gym.