How does it work?
The Corporate Solution runs in 2-3 servers at the customer premises. No interaction with external third parties is needed, and the information is contained at the customer Data Center.
-
Multiformat resources are loaded in the input area and the firstprocess transcribes resources into single format text:
◦ Images / Raster files are OCRed
◦ Speech is transcribed and actors are detected and referenced
◦ PDF, Word and PowerPoint formats are all converted to plain text files.
-
Language Detection and Translation: The language at paragraph level is detected, and if it is not English the text is sent through a neural network specifically designed to translate from the source language into English
-
Specific neural networks receive the monolingual input and produce records with the relevant results (sentiment, importance, referenced entities…) and a graph-model that can later be used to find references in the raw data.
Support services are in charge of managing the collaborative workflow by pipelining the data to and from neural networks and distributing the load to efficiently leverage the hardware and software resources.
Benefits
Sometimes there’s simply too much data and there’s no way the relevant information can be found using traditional resources.
Cost and delivery time are both strong reasons to consider an automated analysis of language resources.
Other Use Cases
Pangeanic delivered a Neural Machine Translation engine wrapped for deployment in the U.S. Department of Defense’s Iron Bank, the secure software repository where applications are hardened and accredited for military use. The engine was trained to handle criminal slang and coded language, supporting crowd monitoring and interrogations with real-time, defense-grade security. This solution helps law enforcement gather intelligence faster, operate with greater transparency, and ensure compliance with strict cybersecurity standards.
