The DCU Discourse Parser for Connective, Argument Identification and Explicit Sense Classification

@inproceedings{Wang2015TheDD,
  title={The DCU Discourse Parser for Connective, Argument Identification and Explicit Sense Classification},
  author={Longyue Wang and Chris Hokamp and Tsuyoshi Okita and Xiaojun Zhang and Qun Liu},
  booktitle={CoNLL},
  year={2015}
}
This paper describes our submission to the CoNLL-2015 shared task on discourse parsing. We factor the pipeline into subcomponents which are then used to form the final sequential architecture. Focusing on achieving good performance when inferring explicit discourse relations, we apply maximum entropy and recurrent neural networks to different sub-tasks such as connective identification, argument extraction, and sense classification. The our final system achieves 16.51%, 12.73% and 11.15… Expand
Finding Arguments as Sequence Labeling in Discourse Parsing
TLDR
The main focus of this effort is to recall cases when Arg1 locates at sentences far before the connective phrase, with some yet limited success. Expand
Window-Based Neural Tagging for Shallow Discourse Argument Labeling
TLDR
This paper describes a novel approach for the task of end-to-end argument labeling in shallow discourse parsing by decomposition of the overall labeling task into subtasks and a general distance-based aggregation procedure. Expand
Semi-Supervised Tri-Training for Explicit Discourse Argument Expansion
TLDR
A novel application of semi-supervision for shallow discourse parsing using a neural approach for sequence tagging and focus on the extraction of explicit discourse arguments to enable the training of large statistical models. Expand
The DCU Discourse Parser: A Sense Classification Task
TLDR
The discourse parsing system developed at Dublin City University for participation in the CoNLL 2015 shared task and especially the sense classification for implicit connectives is described. Expand
Contextualized Embeddings for Connective Disambiguation in Shallow Discourse Parsing
TLDR
A novel model is studied that integrates contextualized word embeddings and predicts whether a connective candidate is part of a discourse relation or not and shows the benefit of training the tasks of connective disambiguation and sense classification together at the same time. Expand
Do We Really Need All Those Rich Linguistic Features? A Neural Network-Based Approach to Implicit Sense Labeling
TLDR
This system extends the two best parsers from previous year's competition by integration of a novel implicit sense labeling component and is grounded on a highly generic, language-independent feedforward neural network architecture incorporating weighted word embeddings for argument spans which obviates the need for (traditional) hand-crafted features. Expand
Argument Labeling of Explicit Discourse Relations using LSTM Neural Networks
TLDR
This paper proposes a Long Short Term Memory (LSTM) based model for argument labeling that learns only from the raw dataset, which is more widely applicable to multiple textual genres and languages. Expand
The CoNLL-2015 Shared Task on Shallow Discourse Parsing
TLDR
The CoNLL-2015 Shared Task is on Shallow Discourse Parsing, a task focusing on identifying individual discourse relations that are present in a natural language text, and the evaluation protocol and metric used during this shared task is presented. Expand
Neural Discourse Modeling
TLDR
A neural discourse parser is proposed which addresses computational and linguistic issues that arise from implicit discourse relation classification in the Penn Discourse Treebank and shifts from the traditional discrete feature paradigm to the distributed continuous feature paradigm, also known as neural network modeling. Expand
Dropped pronoun generation for dialogue machine translation
TLDR
A novel approach to improve the translation of DPs for dialogue machine translation by building a training data for DP generation, and developing a generation model based on recurrent neural networks and language models. Expand
...
1
2
...

References

SHOWING 1-10 OF 12 REFERENCES
Shallow Discourse Parsing with Conditional Random Fields
TLDR
This paper takes a data driven approach to identify arguments of explicit discourse connectives and designs the argument segmentation task as a cascade of decisions based on conditional random fields (CRFs). Expand
A PDTB-styled end-to-end discourse parser
TLDR
This work has designed and developed an end-to-end discourse parser- to-parse free texts in the PDTB style in a fully data-driven approach and significantly improves on the current state-of-the-art connective classifier. Expand
A Unified Framework for Discourse Argument Identification via Shallow Semantic Parsing
TLDR
Evaluation using Penn Discourse Treebank shows that the current research’s parsing approach significantly outperforms the state-of-the-art chunking alternatives. Expand
Comparative Evaluation of Argument Extraction Algorithms in Discourse Relation Parsing
TLDR
It is shown that processing intraand inter-sentential relations separately, reduces the task complexity and significantly outperforms the single model approach. Expand
Global Features for Shallow Discourse Parsing
TLDR
This paper addresses the problem of parsing coherence relations as defined in the Penn Discourse Tree Bank by adopting a two-stage approach where first the local constraints are applied and then global constraints are used on a reduced weighted search space (n-best). Expand
The Penn Discourse TreeBank 2.0
We present the second version of the Penn Discourse Treebank, PDTB-2.0, describing its lexically-grounded annotations of discourse relations and their two abstract object arguments over the 1 millionExpand
Discovering Implicit Discourse Relations Through Brown Cluster Pair Representation and Coreference Patterns
TLDR
Brown cluster pairs are employed to represent discourse relation and coreference patterns are incorporated to identify senses of implicit discourse relations in naturally occurring text to improve baseline performance. Expand
Generative and discriminative algorithms for spoken language understanding
TLDR
Generative and discriminative approaches to modeling the sentence segmentation and concept labeling are studied and it is shown how non-local non-lexical features (e.g. a-priori knowledge) can be modeled with CRF which is the best performing algorithm across tasks. Expand
Investigation of recurrent-neural-network architectures and learning methods for spoken language understanding
TLDR
The results show that on this task, both types of recurrent networks outperform the CRF baseline substantially, and a bi-directional Jordantype network that takes into account both past and future dependencies among slots works best, outperforming a CRFbased baseline by 14% in relative error reduction. Expand
Theano: new features and speed improvements
TLDR
New features and efficiency improvements to Theano are presented, and benchmarks demonstrating Theano's performance relative to Torch7, a recently introduced machine learning library, and to RNNLM, a C++ library targeted at recurrent neural networks. Expand
...
1
2
...