Three-order normalized PMI and other lessons in tensor analysis of verbal selectional preferences
We investigate several questions in transitive verb structure representation by decomposing tensors populated with different subject-verb-object association measures, including a novel generalization of normalized pointwise mutual information to the higher-order (>2) case. Which association measu...
Elmentve itt :
| Szerző: | |
|---|---|
| Testületi szerző: | |
| Dokumentumtípus: | Könyv része |
| Megjelent: |
2022
|
| Sorozat: | Magyar Számítógépes Nyelvészeti Konferencia
18 |
| Kulcsszavak: | Nyelvészet - számítógép alkalmazása |
| Tárgyszavak: | |
| Online Access: | http://acta.bibl.u-szeged.hu/75868 |
| Tartalmi kivonat: | We investigate several questions in transitive verb structure representation by decomposing tensors populated with different subject-verb-object association measures, including a novel generalization of normalized pointwise mutual information to the higher-order (>2) case. Which association measure works the best in modeling verb structures? Should we include occurrences with unfilled arguments in our statistics? We also investigate qualitatively the latent dimensions, and the difference between each noun as a subject versus an object. |
|---|---|
| Terjedelem/Fizikai jellemzők: | 105-120 |
| ISBN: | 978-963-306-848-9 |