The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase 1 project is as follows. Commercially, the technology described herein has the capability to provide faster, granular, dynamic information about safety and security, globally, to firms, universities, governments, and NGOs. This project uses novel methods to improve Natural Language Processing and Machine Learning. As a result, better geo-parsing of digital sources of news about security will result in risk content that can be aggregated, displayed, and analyzed in original ways. This enables security managers at these organizations to better understand risk and protect their staff, providing a higher quality of care. For non-governmental organizations (including firms), this enables improved decision-making about operations, travel, and investment. For governments, this enables improved physical security resource allocation. Socially, this project has the potential to improve transparency and accountability regarding trends about safety and security, by improving the aggregation and visualization of data. As an example, groups of firms and governments in emerging markets can collectively identify previously unnoticed patterns of insecurity, in support of public accountability. This Small Business Innovation Research (SBIR) Phase I project is an innovation over the state of the art in the following ways. First, this project builds on current geo-parsing extraction methodologies by adding methodologies unique to the safety and security space. Second, this project uses external data sources for cross correlations to improve the "aboutness" and granularity of extracted reports. Third, this project exploits contributions from users at the organizational level - as well as individuals. That is, this project supports the growth of an ecosystem in which human users of information also contribute to the quality, volume, and timeliness of that information. This contribution is also intended to improve the geo-parsing methodologies via machine learning. The opportunity is the improvement of geo-parsing extraction mechanisms. The research objectives are to test the hypotheses that NLP algorithms can exploit patterns unique to the safety and security space; that external sources of news can be exploited for improved granularity and "aboutness" scores; and that user-generated content can serve to support an ecosystem of information sharing. The anticipated results are that the above innovations will result in usability scoring sufficient for the safety and security use case.
