| D09-1087 | reported no improvements from | self-training | a PCFG parser on the standard |
| D09-1087 | ubiquitous . Early investigations on | self-training | for parsing have had mixed re |
| D08-1106 | Bootstrapping Bootstrapping ( or | self-training | ) is a general framework for |
| D09-1087 | this section , we discuss how | self-training | is applied to train a PCFG-LA |
| D09-1087 | enhancements in Section 2 , and discuss | self-training | in Section 3 . We then outline |
| D09-1033 | by Ng and Cardie ( 2003 ) in a | self-training | set-up . However , while they |
| D08-1071 | particular , we compare both hints and | self-training | to the two baselines , and then |
| D09-1087 | This is the approach we chose for | self-training | . An alternative approach is |
| D08-1071 | helps significantly , even over | self-training | . We further compare the algorithms |
| D09-1087 | 's parser that the addition of | self-training | data helps the former parser |
| D09-1060 | McClosky et al. ( 2006 ) presented a | self-training | approach for phrase structure |
| D08-1114 | examples of SSL algorithms include | self-training | ( Yarowsky , 1995 ) and co-training |
| D09-1087 | Chinese news articles are used for | self-training | . Since the Chinese parsers in |
| D09-1087 | and alignment edges . <title> | Self-Training | PCFG Grammars with Latent Across |
| D09-1087 | how much does it benefit from | self-training | ? The first question is of special |
| D08-1106 | have described in this paper , | self-training | can be thought of a graph-based |
| D08-1090 | system 's own output , known as | Self-Training | ( Ueffing , 2006 ) has previously |
| D09-1087 | in the BLLIP corpus for English | self-training | . For the Chinese experiments |
| D09-1087 | English and Chi - nese . With | self-training | , a fraction of the WSJ or CTB6 |
| D08-1071 | is relatively ignored . • | Self-training | performance often degrades as |
