Spatial Reference Extraction and Interpretation System Abhishek Amit

Spatial Reference Extraction and Interpretation System Abhishek Amit, Faculty Mentor: Nalini Venkatasubramanian Overview The 911 system is an integral part of emergency response systems. However, it is quickly overloaded in disaster situations. The purpose of this project was to create a computer system to extract location references and other information from 911 calls, and to use this information to help 911 operators process calls more efficiently by giving them a better understanding of the situations they are dealing with. What information did we want to extract? The most important pieces of information in an emergency call are spatial references. Information such as addresses, cities, and phone numbers can be used to map a call to a specific point to which resources can be dispatched. This information is particularly useful for calls from cell phones which do not provide location information. Extracting this information was therefore our first priority. Situational awareness, however, encompasses more than just spatial awareness. Another important component of situational awareness for medical calls is information about the medical emergency, which can be shared with first responders and hospitals who need this information. Some of the data we extracted for medical calls includes: • Victim age • Individual behaviors • Chemical and pharmacologic substances involved • Injuries • Medical signs and symptoms of problems How did we extract this information? By looking for patterns and referencing large lexicons of street and city names in the transcripts of a set of 911 calls, we were able to extract all of the spatial data there was in the calls. Some of the patterns we looked for were: • Numbers followed by a street name, optionally followed by a type such as Street or Drive • City names • Series of seven or ten digits which indicate a phone number We used specialized software designed by the National Institute of Health to extract medical information from medical calls. This software normalized the text and then looked for references to medical terms from a large database. What can we do with the extracted information? The extracted data is stored in a database, and we provide a web-based interface that spatially plots the calls on a map. In addition to this, we developed two query systems that allow users to narrow their focus and find specific events. The first system allows users to search for nearby calls given a location. The second system performs a similarity search by comparing search terms entered by the user with the transcripts of calls in the database. The results from both query systems are used to filter the calls that are displayed on the map, and this can be used to find patterns or other information about calls. How does this help enhance responses to disasters? Automatically extracting this information in real-time from calls opens up many potential applications. Providing 911 operators with a better spatial visualization of all calls happening in a given area can help them spot patterns and duplicates. Going beyond this, we can provide: • Better information sharing with first responders, hospitals, and other affected parties • Detailed analysis tools to study trends in calls, responses to past disasters, and other information Situational awareness is an integral part of disaster response, and this work contributes to this by providing location and medical information extraction from free-text thereby enhancing the 911 system. Acknowledgments Special thanks go to Stuart Ross, Said Shokair, Nalini Venkatasubramanian, Chaitanya Desai, and the entire RESCUE group for their help and contributions to this project. S ummer U ndergraduate 2 R esearch 0 F ellowship in 0 I nformation 6 T echnology aamit@uci. edu www. research. calit 2. net/students/surf-it 2006 www. calit 2. net