The "Corporate Data Extraction" Project Collection 🌱
A collaborative effort aimed at extracting select Green House Gas emission datapoints from sustainability reports in PDF format. The collection currently contains two important projects:
- The "Rule-Based Pipeline" developed by Dr. Ismail Demir (IDS GmbH) and contributors, implementing a PDF text-conversion and graphical table border detection mechanism, with NLP techniques to search for datapoints while associating them with a nearby timestamp.
- The "Model Pipeline" developed by Dr. David Besslich (IDS GmbH) and contributors, implementing a Huggingface Transformer based machine-learning approach for mapping PDF document texts directly to the datapoints.
A web-app demo of these projects is hosted at https://data-extraction-app-aicos-osc-demo.apps.odh-cl2.apps.os-climate.org.
The dockerfiles and their setup may contain mechanisms to allow deployment on Red Hat OpenShift Platform. Refer to Deploying on OpenShift Platform for the respective instructions and how they resulted in a different container setup.