A Discriminative Latent Variable Model for Statistical Machine Translation.
Large-scale discriminative machine translation promises to further the state-of-the-art, but has failed to deliver convincing gains over current heuristic frequency count systems.We argue that a principle reason for this failure is not dealing with multiple, equivalent translations. We present a tra...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference item |
| Published: |
Association for Computer Linguistics
2008
|
| _version_ | 1797066854339969024 |
|---|---|
| author | Blunsom, P Cohn, T Osborne, M |
| author2 | McKeown, K |
| author_facet | McKeown, K Blunsom, P Cohn, T Osborne, M |
| author_sort | Blunsom, P |
| collection | OXFORD |
| description | Large-scale discriminative machine translation promises to further the state-of-the-art, but has failed to deliver convincing gains over current heuristic frequency count systems.We argue that a principle reason for this failure is not dealing with multiple, equivalent translations. We present a translation model which models derivations as a latent variable, in both training and decoding, and is fully discriminative and globally optimised. Results show that accounting for multiple derivations does indeed improve performance. Additionally, we show that regularisation is essential for maximum conditional likelihood models in order to avoid degenerate solutions. © 2008 Association for Computational Linguistics. |
| first_indexed | 2024-03-06T21:47:59Z |
| format | Conference item |
| id | oxford-uuid:4a44c3a3-e958-4625-8f6f-d49e6b2c4d82 |
| institution | University of Oxford |
| last_indexed | 2024-03-06T21:47:59Z |
| publishDate | 2008 |
| publisher | Association for Computer Linguistics |
| record_format | dspace |
| spelling | oxford-uuid:4a44c3a3-e958-4625-8f6f-d49e6b2c4d822022-03-26T15:36:28ZA Discriminative Latent Variable Model for Statistical Machine Translation.Conference itemhttp://purl.org/coar/resource_type/c_5794uuid:4a44c3a3-e958-4625-8f6f-d49e6b2c4d82Symplectic Elements at OxfordAssociation for Computer Linguistics2008Blunsom, PCohn, TOsborne, MMcKeown, KMoore, JTeufel, SAllan, JFurui, SLarge-scale discriminative machine translation promises to further the state-of-the-art, but has failed to deliver convincing gains over current heuristic frequency count systems.We argue that a principle reason for this failure is not dealing with multiple, equivalent translations. We present a translation model which models derivations as a latent variable, in both training and decoding, and is fully discriminative and globally optimised. Results show that accounting for multiple derivations does indeed improve performance. Additionally, we show that regularisation is essential for maximum conditional likelihood models in order to avoid degenerate solutions. © 2008 Association for Computational Linguistics. |
| spellingShingle | Blunsom, P Cohn, T Osborne, M A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title | A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title_full | A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title_fullStr | A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title_full_unstemmed | A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title_short | A Discriminative Latent Variable Model for Statistical Machine Translation. |
| title_sort | discriminative latent variable model for statistical machine translation |
| work_keys_str_mv | AT blunsomp adiscriminativelatentvariablemodelforstatisticalmachinetranslation AT cohnt adiscriminativelatentvariablemodelforstatisticalmachinetranslation AT osbornem adiscriminativelatentvariablemodelforstatisticalmachinetranslation AT blunsomp discriminativelatentvariablemodelforstatisticalmachinetranslation AT cohnt discriminativelatentvariablemodelforstatisticalmachinetranslation AT osbornem discriminativelatentvariablemodelforstatisticalmachinetranslation |