• Corpus ID: 15813792

Learning to Search for Dependencies

@article{Chang2015LearningTS,
  title={Learning to Search for Dependencies},
  author={Kai-Wei Chang and He He and Hal Daum{\'e} and John Langford},
  journal={ArXiv},
  year={2015},
  volume={abs/1503.05615}
}
We demonstrate that a dependency parser can be built using a credit assignment compiler which removes the burden of worrying about low-level machine learning details from the parser implementation. The result is a simple parser which robustly applies to many languages that provides similar statistical and computational performance with best-to-date transition-based parsing approaches, while avoiding various downsides including randomization, extra feature requirements, and custom learning… 

Figures and Tables from this paper

A Credit Assignment Compiler for Joint Prediction
TLDR
It is shown the search space can be defined by an arbitrary imperative program, turning learning to search into a credit assignment compiler, and algorithmic improvements for the compiler are made to radically reduce the complexity of programming and the running time.
Efficient programmable learning to search
TLDR
It is shown that the search space can be defined by an arbitrary imperative program, reducing the number of lines of code required to develop new structured prediction tasks by orders of magnitude.
Learning to Prune: Exploring the Frontier of Fast and Accurate Parsing
Pruning hypotheses during dynamic programming is commonly used to speed up inference in settings such as parsing. Unlike prior work, we train a pruning policy under an objective that measures
Improving Coreference Resolution by Learning Entity-Level Distributed Representations
TLDR
A neural network based coreference system that produces high-dimensional vector representations for pairs of coreference clusters that learns when combining clusters is desirable and substantially outperforms the current state of the art on the English and Chinese portions of the CoNLL 2012 Shared Task dataset.
Sanaphor++: Combining Deep Neural Networks with Semantics for Coreference Resolution
TLDR
The goal of Sanaphor++ is to improve the clustering part of the coreference resolution in order to know if two clusters have to be merged or not once the pairs of mentions have been identified.
Knowledge Extraction in Web Media: At The Frontier of NLP, Machine Learning and Semantics
TLDR
This research presents a preliminary framework based on a novel hybrid architecture for an entity linking system, that combines methods from the Natural Language Processing (NLP), information retrieval and semantic fields, and proposes a modular approach in order to be as independent as possible of the text to be processed.
Dynamic Feature Induction: The Last Gist to the State-of-the-Art
TLDR
A novel technique called dynamic feature induction is introduced that keeps inducing high dimensional features automatically until the feature space becomes ‘more’ linearly separable, and shows the state-of-the-art results for both tasks.
A Neural Entity Coreference Resolution Review
The UMD Neural Machine Translation Systems at WMT17 Bandit Learning Task
TLDR
A standard neural machine translation system is built and extended in two ways: (1) robust reinforcement learning techniques to learn effectively from the bandit feedback, and (2) domain adaptation using data selection from a large corpus of parallel data.
Learning to Search for Recognizing Named Entities in Twitter
TLDR
This work presented in this work their participation in the 2nd Named Entity Recognition for Twitter shared task, and showed that LOD features improved substantially the performance of their system as they counter-balance the lack of context in tweets.
...
...

References

SHOWING 1-10 OF 39 REFERENCES
A Tabular Method for Dynamic Oracles in Transition-Based Parsing
We develop parsing oracles for two transition-based dependency parsers, including the arc-standard parser, solving a problem that was left open in (Goldberg and Nivre, 2013). We experimentally show
Online Large-Margin Training of Dependency Parsers
TLDR
An effective training algorithm for linearly-scored dependency parsers that implements online large-margin multi-class training on top of efficient parsing techniques for dependency trees is presented.
Dynamic Feature Selection for Dependency Parsing
TLDR
This work proposes a faster framework of dynamic feature selection, where features are added sequentially as needed, edges are pruned early, and decisions are made online for each sentence.
A Fast and Accurate Dependency Parser using Neural Networks
TLDR
This work proposes a novel way of learning a neural network classifier for use in a greedy, transition-based dependency parser that can work very fast, while achieving an about 2% improvement in unlabeled and labeled attachment scores on both English and Chinese datasets.
Dynamic Programming Algorithms for Transition-Based Dependency Parsers
We develop a general dynamic programming technique for the tabulation of transition-based dependency parsers, and apply it to obtain novel, polynomial-time algorithms for parsing with the
Statistical Dependency Analysis with Support Vector Machines
TLDR
Though the result is little worse than the most up-to-date phrase structure based parsers, it looks satisfactorily accurate considering that the parser uses no information from phrase structures.
Transition-based Dependency Parsing with Rich Non-local Features
TLDR
This paper shows that it can improve the accuracy of transition-based dependency parsers by considering even richer feature sets than those employed in previous systems by improving the accuracy in the standard Penn Treebank setup and rivaling the best results overall.
Incrementality in Deterministic Dependency Parsing
TLDR
It is concluded that strict incrementality is not achievable within this framework and it is shown that it is possible to minimize the number of structures that require non-incremental processing by choosing an optimal parsing algorithm.
Simple Semi-supervised Dependency Parsing
TLDR
This work focuses on the problem of lexical representation, introducing features that incorporate word clusters derived from a large unannotated corpus, and shows that the cluster-based features yield substantial gains in performance across a wide range of conditions.
A Credit Assignment Compiler for Joint Prediction
TLDR
It is shown the search space can be defined by an arbitrary imperative program, turning learning to search into a credit assignment compiler, and algorithmic improvements for the compiler are made to radically reduce the complexity of programming and the running time.
...
...