Leveraging Dynamic Graph Word Embedding for Efficient Contextual Representations
Published in International Symposium on Information and Communication Technology, 2025
Proposes creating word embeddings from a text by first mapping to a graph to capture token relationships. Shows that the GNN ARMA-Conv can outperform sequential based machine learning models for text classification.
Recommended citation: Himes, R.E., Tran, HA., Tran, T.X. (2025). Leveraging Dynamic Graph Word Embedding for Efficient Contextual Representations. In: Buntine, W., Fjeld, M., Tran, T., Tran, MT., Huynh Thi Thanh, B., Miyoshi, T. (eds) Information and Communication Technology. SOICT 2024. Communications in Computer and Information Science, vol 2352. Springer, Singapore. https://doi.org/10.1007/978-981-96-4288-5_20
Download Paper