“…First, compared to purely metric-based techniques, considers the contextual semantics in addition to the word meanings themselves. Specifically, purely metric-based text embeddings typically assign weights to individual n-grams with minimal consideration of the semantic meaning and relationships between the words (Kasumba & Neumann, 2022; van Tussenbroek, 2020). Second, is better able to handle unseen words (Athiwaratkun et al, 2018; Won et al, 2021).…”