Skip to main content
Back to How We Work

Understanding and Extraction

Synergy Group was engaged to develop a solution involving automation and machine learning platforms to test the feasibility of ingesting, processing and extracting targeted information from semi and unstructured legacy maintenance documentation, for future data storage, uplift and interrogation.

Defence Industry
5 minutes
What did we set out to do?

The automation team (as a partnered group between Synergy and Reveal Group) was tasked with an innovation project which sought to test the capability of currently available automation and artificial intelligence/machine learning platforms to perform targeted data extraction on legacy documentation. 

The task was extremely dynamic, being derived from an innovation program attempting to understand current technology limits. As a result the scope was extremely flexible, but the expectations of the solution were extremely high. 

Specifically, the intent was for a fully functioning solution to be able to ingest (read) any set of legacy maintenance documentation and identify a set number of information requirements; for example does the document include part numbers, safety warnings, load limits etc. 

Within the project scope and timeframe, a subset of documentation was targeted for testing. The preferred solution preferably need to function within the existing Defence ICT infrastructure.  


What was achieved?

The team reviewed and tested a large range of market products, before short listing only a handful of candidates deemed technically viable. When combined with Defence ICT requirements a combination of Blue Prism and WATSON Knowledge Studio elements were selected for solution design and testing. 

The final product demonstrated that the task was achievable with limitations. The WATSON suite was capable of successfully ingesting and reliably reading the test set of documentation, ultimately producing a raw data output of the required information. The Blue Prism solution component were then capable of manipulating the raw data (in JSON form) to identify data components requiring human review and validation. After validation, the Blue Prism solution could produce a final data output for future use. 

However, the WATSON model would require modification as the data set was redirected in order to maintain its accuracy between radically different document sets.  

The organisation innovation program was able to prove a solution concept which both reviewed and tested the leading market capabilities for automation and machine learning, for application to a real-time organisational issue. The solution is under considerations for future implementation. 

Back to How We Work
Share Article