The expert modeling system provides tools to develop a Subject Matter Expert matching system, from an attributed corpus, based on the Author-Topic-Model. This was a prototype developed for NASA's Jet Propultion lab that matches tickets describing late stage mission critical anomalies with best candidate subject matter experts.
For mission critical anomalies, described as engineering faults that effect late stage and deployed missions, an engineering ticket is opened. In the body of this ticket there is a free-text description of the anomaly occuring. The process of matching an expert, to solve a ticket, could take several weeks and was usually handled by word of mouth and monopolizing on tribal knowlege.
This program includes the pre-processing, processing, model building, and prototyped information retrieval system that automates this expensive and faulty HR task. Given as few as 30 words, from the ticket description (which are usually 600+) we are able to match the correct engineer to close tickets within the first 15 suggestions.
The program is located in the ./code
directory, including notes on how to configure the
model for training and hosting to a local site.