Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Cross-Lingual Word Sense Disambiguation (CLWSD), which seeks to distinguish various senses of a word in one language by utilizing information from another language, presents a potential approach for WSD tasks. This study delves into the impact of noun-verb differences on the effectiveness of CLWSD. We began by creating sense inventories using the CLWSD approach, extracting translations of a target word from an English-Chinese corpus. Human verification was then carried out to select the final class labels and training as well as testing instances. Subsequently, features such as unigrams, POS tags, domain information and word embeddings were extracted and converted into machine-readable data. These data were employed to train classifiers for 10 selected polysemous nouns and 10 selected polysemous verbs, employing Support Vector Machines (SVM). The findings revealed that linguistic features, such as lexical characteristics, syntactic roles, and word embeddings, influenced the performance of CLWSD models differently for nouns and verbs. While lexical features contributed to the enhancement of noun sense disambiguation, syntactic features played important roles in verb sense disambiguation. Moreover, word embeddings increased the performance of both noun and verb disambiguation. The outcomes of this study yielded valuable linguistic insights for researchers aiming to provide effective solutions for verb and noun sense disambiguation.