Self-Training PCFG Grammars with Latent Annotations Across Languages

By Zhongqiang Huang, Mary P. Harper. 2009

104 citations recorded by the source corpus.

We investigate the effectiveness of self-training PCFG grammars with latent annotations (PCFG-LA) for parsing languages with different amounts of labeled training data. Compared to Charniak's lexicalized parser, the PCFG-LA parser was more effectively adapted to a language for which parsing has been less well developed (i.e., Chinese) and benefited more from self-training. We show for the first time that self-training is able to significantly improve the performance of the PCFG-LA parser, a single generative parser, on both small and large amounts of labeled training data. Our approach achieves state-of-the-art parsing accuracies for a single parser on both English (91.5%) and Chinese (85.2%).

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Recurrent Neural Network Grammars (2016, neutral): “2 Huang and Harper (2009) D 84.”
Grammar as a Foreign Language (2014, methodology): “We include our reimplementation of BerkeleyParser as an additional baseline, and also add a self-training experiment where it is trained on its own output, similar to Huang & Harper (2009).”
Exploiting Cloze-Questions for Few-Shot Text Classification and Natural Language Inference (2020, background): “, 2015), parsing (McClosky et al., 2006; Reichart and Rappoport, 2007; Huang and Harper, 2009), machine translation (Hoang et al.”
Revisiting Self-Training for Neural Sequence Generation (2019, methodology): “In the ﬁeld of natural language processing, some early work have successfully applied self-training to word sense disambiguation (Yarowsky, 1995) and parsing (McClosky et al., 2006; Reichart & Rappoport, 2007; Huang & Harper, 2009).”
slimIPL: Language-Model-Free Iterative Pseudo-Labeling (2020, background): “Many recent works on self-training propose and validate these extensions, including those in computer vision [12, 13], natural language processing [14, 15, 16, 17, 18, 19, 20], ASR [21, 7, 22, 8, 9], and speech translation [23].”
Word Segmentation, Unknown-word Resolution, and Morphological Agreement in a Hebrew Parsing System (2013, background): “In Chinese, such separation is not available.”
Self-Training with Products of Latent Variable Grammars (2010, methodology): “Huang and Harper (2009) argued that the additional self-labeled training data adds a smoothing effect to the grammars, supporting an increase in model complexity without overﬁtting.”
Ambiguity-aware Ensemble Training for Semi-supervised Dependency Parsing (2014, methodology): “For unlabeled data, we follow Chen et al. (2013) and use the BLLIP WSJ corpus (Charniak et al., 2000) for English and Xinhua portion of Chinese Gigaword Version 2.0 (LDC2009T14) (Huang, 2009) for Chinese.”
Embedding epistemic modals in English: A corpus-based study (2012, unknown): “For details on the general accuracy of the parser, see Huang & Harper 2009.”
Joint POS Tagging and Transition-based Constituent Parsing in Chinese with Non-local Features (2014, background): “Table 5 shows the parsing performances on CTB6.”
Products of Random Latent Variable Grammars (2010, background): “S E L F Huang and Harper (2009) 91.”
Structured Training for Neural Network Transition-Based Parsing (2015, background): “(2006) and Huang and Harper (2009) showed that iteratively re-training a single model (“self-training”) can be used to improve parsers in certain settings; Petrov et al. (2010) built on this work and showed that a slow and accurate parser can be used to “up-train” a faster but less accurate parser.”