Latent Dirichlet Allocation-based method and Cosine Similarity: Aligning Research Publications with the United Nations’ Sustainable Development Goals
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3033234Utgivelsesdato
2022Metadata
Vis full innførselSamlinger
- Master of Science [1822]
Sammendrag
According to the United Nations Statistics Division (2021) the development of the
Sustainable Development Goals is running behind schedule. Higher education plays a
significant role in prioritising and implementing research as part of their sustainability
agenda. Although there are compelling reasons for research faculty to focus research
contributions on advancing the Sustainable Development Goals, current applications
for identifying such efforts heavily rely on manual involvement.
In order to assess how the research publications at a Norwegian business
school were aligned with the Sustainable Development Goals, we investigated and
implemented an SDG classifier based on Latent Dirichlet Allocation developed by
LaFleur (2019). In addition, cosine similarity was used as an alternative method for
identifying the most similar research publications in the corpus to the Sustainable
Development Goals.
Our results show that BI Norwegian Business School, has made clear
contributions to the Sustainable Development Goals, when using both the SDG
classifier and cosine similarity. As a result, both methods are adequate for identifying
Sustainable Development Goals in research publications for our purposes. However,
the SDG classifier produces more reliable results than cosine similarity as it is able
to capture Sustainable Development Goals-related topics in research publications
where they are not always prominent.
Beskrivelse
Masteroppgave(MSc) in Master of Science in Business Analytics - Handelshøyskolen BI, 2022