Translation-based Word Sense Disambiguation: Appendices

Abstract

Appendix 1: Norwegian ambiguous lemmas in the ENPC where at least two senses have a frequency greater than or equal to 10. Appendix 2: English ambiguous lemmas in the ENPC where at least two senses have a frequency greater than or equal to 10. Appendix 3: Results from model selection (cross-validation) in Chapter 9 with knowledge source=WORDS. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected). Appendix 4: Results from model selection (cross-validation) in Chapter 9 with knowledge source=SEMANTIC-FEATURES. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected). Appendix 5: Results from model selection (cross-validation) in Chapter 9 with knowledge source=RELATED-WORDS. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected). Appendix 6: Results from model selection (cross-validation) in Chapter 10 with knowledge source=WORDS. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected). Appendix 7: Results from model selection (cross-validation) in Chapter 10 with knowledge source=SEMANTIC-FEATURES. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected). Appendix 8: Results from model selection (cross-validation) in Chapter 10 with knowledge source=RELATED-WORDS. The target words are ordered alphabetically. Evaluated with 5-fold cross validation and Overall Accuracy (measured as total recall). The best accuracy in each group is marked in bold-face (in case of ties, the model with the smallest context window is selected).