Abstract
In natural language generation using symbolic grammars, state-of-the-art realisation rankers use statistical models incorporating both language model and structural features. The rankers depend on multiple structures produced by the particular large-scale symbolic grammars to rank the output; for languages with smaller resources and in-development grammars, we look at the feasibility of an alternative source of structural features, unsupervised parsers. We show that, in spite of their lower quality of structure, raw sets of unsupervised parse features can be helpful with smaller language models; and that the parses do contain particular elements that can be highly useful, improving performance on our classification task by up to 10% on 60% of the test set leading to an overall improvement under a back-off model.
| Language | English |
|---|---|
| Title of host publication | 24th International Conference on Computational Linguistics |
| Subtitle of host publication | Proceedings of COLING 2012: Technical Papers |
| Editors | Martin Kay, Christian Boitet |
| Place of Publication | Mumbai |
| Publisher | Indian Institute of Technology |
| Pages | 1811-1830 |
| Number of pages | 20 |
| Publication status | Published - 2012 |
| Event | 24th International Conference on Computational Linguistics, COLING 2012 - Mumbai, India Duration: 8 Dec 2012 → 15 Dec 2012 |
Other
| Other | 24th International Conference on Computational Linguistics, COLING 2012 |
|---|---|
| Country | India |
| City | Mumbai |
| Period | 8/12/12 → 15/12/12 |
Fingerprint
Cite this
}
Is bad structure better than no structure? unsupervised parsing for realisation ranking. / Motazedi, Yasaman; Dras, Mark; Lareau, François.
24th International Conference on Computational Linguistics: Proceedings of COLING 2012: Technical Papers. ed. / Martin Kay; Christian Boitet. Mumbai : Indian Institute of Technology, 2012. p. 1811-1830.Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding contribution › Research › peer-review
TY - GEN
T1 - Is bad structure better than no structure?
T2 - unsupervised parsing for realisation ranking
AU - Motazedi, Yasaman
AU - Dras, Mark
AU - Lareau, François
PY - 2012
Y1 - 2012
N2 - In natural language generation using symbolic grammars, state-of-the-art realisation rankers use statistical models incorporating both language model and structural features. The rankers depend on multiple structures produced by the particular large-scale symbolic grammars to rank the output; for languages with smaller resources and in-development grammars, we look at the feasibility of an alternative source of structural features, unsupervised parsers. We show that, in spite of their lower quality of structure, raw sets of unsupervised parse features can be helpful with smaller language models; and that the parses do contain particular elements that can be highly useful, improving performance on our classification task by up to 10% on 60% of the test set leading to an overall improvement under a back-off model.
AB - In natural language generation using symbolic grammars, state-of-the-art realisation rankers use statistical models incorporating both language model and structural features. The rankers depend on multiple structures produced by the particular large-scale symbolic grammars to rank the output; for languages with smaller resources and in-development grammars, we look at the feasibility of an alternative source of structural features, unsupervised parsers. We show that, in spite of their lower quality of structure, raw sets of unsupervised parse features can be helpful with smaller language models; and that the parses do contain particular elements that can be highly useful, improving performance on our classification task by up to 10% on 60% of the test set leading to an overall improvement under a back-off model.
UR - http://www.scopus.com/inward/record.url?scp=84876789571&partnerID=8YFLogxK
M3 - Conference proceeding contribution
SP - 1811
EP - 1830
BT - 24th International Conference on Computational Linguistics
A2 - Kay, Martin
A2 - Boitet, Christian
PB - Indian Institute of Technology
CY - Mumbai
ER -