Motivation and Task:

Companies are increasingly being forced to meet government regulations and customer demands pertaining to sustainable production and sustainable product offerings. As result, companies must find, adopt and develop new methods of production and consumption to keep their license to operate. One emerging and expanding knowledge field to search for relevant best practices is the “circular economy”, which refers to a broad model of production and consumption that involves sharing, leasing, reusing, repairing, refurbishing and recycling existing materials and products as long as possible.

Within the knowledge field of the “circular economy”, numerous solutions are already being proposed and described in theory or as best practice already adopted and implemented by companies. The challenge stems from the fact that this knowledge is dispersed over many sources, unstructured, heterogeneuous in format and detail, and therefore very difficult to search systematically for practicioners looking for suitable solutions ot at least inspirations.

The goal of this thesis is to develop and implement an approach that makes the knowledge field “circular economy” more accessible and searchable by deploying tools from natural language processing and machine learning.

Possible subtasks (not all required for one thesis):

• Extracting text data on the topic of circular economy:
  • Dedicated websites / best practice databases
  • Scientific articles
  • (Social) media & press articles
  • Company reports
• Concepts / keyphrase extraction
• Named entities (firms, industries) extraction
• Constructing an ontology / knowledge graph
• Generating knowledge graph embeddings
• Knowledge graph analysis and visualization
• Implement keyword searches over the knowledge graph

Expectations:

• Master students with an interest in circular economy
• Prior experience in Python / R
• Prior experience with or strong motivation to learn about NLP
• Thesis can be written in German or English
• Thesis can be assigned to several students dealing with different data sources