TACTILE: Transparent And Customizable Textual Information Extraction from Large Textual Sources

The ability to analyze the information contained in large textual repositories is critical for data-driven decision making and mission success. Mission owners must be accountable for their decisions and that is why any data supporting a decision must be traced to its original source. Minsky is a no-code ML/NLP platform for non-technical users that can be used to efficiently and effectively analyze vast sources of textual data. The platform is highly customizable and it provides transparent access to the information contained in those sources. With our platform SMEs can train traceable custom ML models in hours, not months and apply them to track events and  monitor risks and opportunities across entities of interest. The work planned in this proposal will improve our transparent semi-supervised data driven ML techniques to further optimize Minsky customization capabilities. Furthermore, the proposed improvements will ensure the adaptability of the platform to dynamic environments. As a result mission owners will be able to customize the platform towards their specific information needs using the most relevant data sources with minimal effort and to continuously adapt the models to a constantly changing environment.