Improving Search Accessibility at the South African SDG Hub using Semantic Matching and Topic Modelling.
Jesse Parvess , Shaun Johnson
Abstract: To improve access to research that is pertinent to the sustainable development goals (SDGs), that was enacted by the United Nations in 2015, the South African SDG Hub created a metadatabase of relevant articles. For this objective to be effective the policymaker must be able to query the metadatabase of papers with a search query that returns relevant research papers. We improved this search functionality using contextual word embeddings (transformers). Additionally the policymaker will benefit from richer information representations of the search results. To further solve the high-level overview problem of finding papers of interest (opposed to finding a needle in a haystack), subtopics within the search results are available. This topic capability will potentially unlock quicker understandings, or simply empower policymakers with filtering options over the already contextually relevant search results. To do this, we used LDA topic analysis.