INITIAL SITUATION

At the customer, a German car manufacturer, mechanics were tasked with collecting repair information by manually writing reports.

Going over the reports and combining the information to create insight was tedious and inefficient.

SOLUTION

An ingestion mechanism including OCR digitization, artefact removal and text cleansing was created and has since been fed well over a million reports.

As a final step in the Natural Language Processing pipeline, topic modelling was used to tag the documents according to their content.

Furthermore, a web UI was built for efficient retrieval and exploration of the parsed documents.

BUSINESS VALUE

It was much easier to understand which problems were occuring, and group them together into actionable clusters.

This allowed much quicker reaction and more robust failure prevention.

FRAMEWORK & TOOLS

Apache Solr and Spark, and Banana Framework.