Muti-Path Shift-Reduce Parsing with Online Training
نویسندگان
چکیده
In this paper we describe an enhanced shift-reduce parsing method which differs from the traditional transition-based model in two ways: first, we maintain multiple transition paths after each transition step, in order to alleviate the serious risk of going astray for the only-one-path transition; second, we adopt the online training algorithm rather than the classical training-after-extraction method, to obtain more robust discriminativity of the classifier for transition decision. Experiments on the Tsinghua Chinese Treebank show that the enhanced model gains obvious improvement over the baseline. And in the CIPS evaluation task, a neat implementation without tricks could achieved nearly the state-of-the-art performance. Keywords: Parsing, Transition-based, Online training Äu3Ôö õ ́£?8 é{©Û . ñ©R,=Ê,4+ ¥IÆ OEâïĤU&E?n:¢ ¿ {jiangwenbin, xionghao, liuqun}@ict.ac.cn Á : ©£ã «Or. £?8 é{©Û ."DÚ £?8 {'§kü U?μ1§ T .ÏL3z=£Ú 3õG ¢yõ ́G =£§ ü ́»£?8 wÍü$ G =£ ØǶ 1 § DÚ 3Ä ¢~þÔö©aì {§T .æ^3Ôö aN!©aìëê§ lÔö a õ ©ÛG l S ° ûüUå"3uä¥þ ¢ y¢§ùüU?üÑwÍJp ©Û5 U" +· XÚØ/Ï?Û[E|§ E3CIPSμÿ¥ * ¤1" ' :é{©Û,£?8 ,3Ôö
منابع مشابه
Online Learning for Deterministic Dependency Parsing
Deterministic parsing has emerged as an effective alternative for complex parsing algorithms which search the entire search space to get the best probable parse tree. In this paper, we present an online large margin based training framework for deterministic parsing using Nivre’s Shift-Reduce parsing algorithm. Online training facilitates the use of high dimensional features without creating me...
متن کاملExpected F-Measure Training for Shift-Reduce Parsing with Recurrent Neural Networks
We present expected F-measure training for shift-reduce parsing with RNNs, which enables the learning of a global parsing model optimized for sentence-level F1. We apply the model to CCG parsing, where it improves over a strong greedy RNN baseline, by 1.47% F1, yielding state-of-the-art results for shiftreduce CCG parsing.
متن کاملExploring beam-based shift-reduce dependency parsing with DyALog: Results from the SPMRL 2013 shared task
The SPMRL 2013 shared task was the opportunity to develop and test, with promising results, a simple beam-based shift-reduce dependency parser on top of the tabular logic programming system DYALOG. The parser was also extended to handle ambiguous word lattices, with almost no loss w.r.t. disambiguated input, thanks to specific training, use of oracle segmentation, and large beams. We believe th...
متن کاملA Shift-Reduce Parsing Algorithm for Phrase-based String-to-Dependency Translation
We introduce a shift-reduce parsing algorithm for phrase-based string-todependency translation. As the algorithm generates dependency trees for partial translations left-to-right in decoding, it allows for efficient integration of both n-gram and dependency language models. To resolve conflicts in shift-reduce parsing, we propose a maximum entropy model trained on the derivation graph of traini...
متن کاملExploiting Language Variants Via Grammar Parsing Having Morphologically Rich Information
In this paper, the development and evaluation of the Urdu parser is presented along with the comparison of existing resources for the language variants Urdu/Hindi. This parser was given a linguistically rich grammar extracted from a treebank. This context free grammar with sufficient encoded information is comparable with the state of the art parsing requirements for morphologically rich and cl...
متن کامل