Finding values in green hydrogen using topic modelling

Abstract

Increased pace of developments strain the ability of policy makers to be timely and sufficiently informed. While there are already sufficient methods available for gauging what plays a role, topic modelling is a novel method that has the potential to be deployed at high speed with low effort. Values play an important role as it shapes policy making and in turn affects stakeholders. A semi-supervised topic modelling method called correlation explanation (CorEx) was used for this purpose as it allows steering the model to find aforementioned values. Green hydrogen is used as a case study where topic modelling is used to find values in scientific literature on this subject. Green hydrogen is envisioned to play a more important role in the context of decarbonisation. Such a transition has a major influence on society engaging business and households alike. Three different value sets are used reflecting different perspectives. These perspectives focus on corporate values, public values and the context in which hydrogen is discussed respectively. It was found that using topic modelling to identify values is highly constrained by its given inputs and processed outputs however. A methodological framework is therefore proposed. It suggests how topic modelling can be conducted for the purpose of identifying values playing a role in a certain domain. This framework consists of five components which are value definition, corpus selection, language processing, topic modelling and result interpretation. Utilising this framework helps with structuring the topic modelling procedure and identify bottlenecks in result quality.