Name
Abstract | ||
|---|---|---|
We show that the automatically induced latent variable grammars of Petrov et al. (2006) vary widely in their underlying representations, depending on their EM initialization point. We use this to our advantage, combining multiple automatically learned grammars into an unweighted product model, which gives significantly improved performance over state-of-the-art individual grammars. In our model, the probability of a constituent is estimated as a product of posteriors obtained from multiple grammars that differ only in the random seed used for initialization, without any learning or tuning of combination weights. Despite its simplicity, a product of eight automatically learned grammars improves parsing accuracy from 90.2% to 91.8% on English, and from 80.3% to 84.5% on German. |
Year | Venue | Keywords |
|---|---|---|
2010 | HLT-NAACL | unweighted product model,improved performance,induced latent variable grammar,multiple grammar,random latent variable grammar,random seed,em initialization point,combination weight,underlying representation,state-of-the-art individual grammar,latent variable |
Field | DocType | ISBN |
Rule-based machine translation,Product model,Computer science,Latent variable,Speech recognition,Natural language processing,Artificial intelligence,Parsing,Initialization,Random seed,Machine learning | Conference | 1-932432-65-5 |
Citations | PageRank | References |
30 | 1.31 | 32 |
Authors | ||
1 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Slav Petrov | 1 | 2405 | 107.56 |