FlyBy
Many large agencies are responsible for handling incoming data and reports daily from sources nationwide. These agencies must enter and store far more data than any single analyst can process in a reasonable amount of time. Such a deluge of data may lead many analysts to miss critical information, leading to employee burnout and low morale. Topologe created FlyBy to cluster data using NLP techniques of measuring text similarity, so that messages with similar words, phrases, and topics can be grouped together in 3-D space. After clustering, a human analyst can use the tool to inspect a cluster, determine the common threads of the cluster, and process the group of messages as a single entity. Once identified, clusters may be packaged and sent to the relevant maintenance group for further analysis. This approach takes tens of thousands of messages and sorts them into a few dozen clusters, reducing the processing time from weeks to hours for a single analyst.
Philippines Agricultural Consumption Evaluator (PACE)
In order to combat shortages in crop yields in the near future, the Philippines' Department of Agriculture must understand how much monthly production they can expect from their farms, how much rice they must import from outside markets, and what cost they will pay to meet their national demand. Using a combination of time series forecasting and statistical modelling, Philippines Agricultural Consumption Evaluator (PACE) is a multi-use tool that pinpoints which "levers" that the government can manipulate to maximize production, estimates of the next 12 months of neighboring competitive import prices, and an interactive chart displaying cumulative rice stocks based on imports. Understanding how economic taxes and fees affect crop yield help the government will empower the government to maximize production, while the price model will enable better ability to reduce import cost by selecting cheaper time frames to import rice while keeping a safe stock for citizens in case of emergency.
Air Traffic Communications Visualizer
The FAA has worked for decades to enhance the safety and predictability of aviation traffic in the United States. This success is supported by the diligent planning and coordination of aircraft movements by pilots and air traffic controllers. While the rate of accidents in U.S. aviation is very low, there is always room for improvement. This tool processes transcripts of ATC communications with pilots to extract relevant insights, identifying the actions pilots are instructed to take and those they intend to take. The data serves as an additional layer of validation and verification to ensure that spoken instructions are accurately understood by both pilots and controllers. It can also help identify imminent risks, such as runway contention or in-air collisions.
CyBy - Flyby for Cybersecurity
By applying lessons learned within cybersecurity threat detection to 3-D space, Topologe created an adaptable version of FlyBy capable of identifying malicious traffic, examining patterns and trends within clusters, and triaging threats. With the naked eye, analysts can assess incoming traffic, easily sequester clusters of threats, and discard large volumes of harmless network traffic. The more users interact with FlyBy, the more the AI learns from human experts which threat signals to monitor over time.
Remaining Useful Life
In the transportation industry, machine learning innovations are paving the way for predictive and preventive maintenance that can allow aircraft technicians to repair vehicles when they require maintenance rather than on a traditional fixed schedule. After measuring various health indicators in aircraft engines over their life cycles, Topologe developed an engine health predictor that provides technicians with a health score, saving valuable diagnostic time in the hangar. This preventative technology can also be applied to facilities management (e.g. servers, network equipment, radars, runway lights, etc).
Sentiment Analysis
Utilizing Natural Language Processing (NLP) and social media mining, our Sentiment Analysis model provides a comprehensive analysis of large data sets to understand attitudes. Filtering the body of posts by dates, sources, and sentiment levels produces a method of tracking sentiments across time and social media sources to provide more insight on public sentiment of a given organization across multiple categories of interest. This search engine can be used to automate analysis for a variety of linguistic settings, such as reviews made by employees or comments left by technicians and manufacturers. In addition, NLP can perform text summarization, root cause analysis, detect disinformation, etc.