Word Usage Graph

Word usage graphs (WUGs) represent relationships between words, often derived from co-occurrence patterns or semantic embeddings, to model textual meaning and facilitate various NLP tasks. Current research focuses on enriching WUGs with semantic information, such as cluster definitions or sense labels, and developing efficient algorithms, including graph neural networks and transformer-based models, for graph construction, manipulation, and analysis. These advancements improve the accuracy and efficiency of applications like keyphrase extraction, semantic role labeling, and even cross-lingual form understanding, ultimately enhancing the interpretability and performance of NLP systems.