Stanford’s Graph-based Neural Dependency Parser at the CoNLL 2017 Shared Task

By Timothy Dozat, Peng Qi, Christopher D. Manning. 2017

234 citations recorded by the source corpus.

This paper describes the neural dependency parser submitted by Stanford to the CoNLL 2017 Shared Task on parsing Universal Dependencies. Our system uses relatively simple LSTM networks to produce part of speech tags and labeled dependency parses from segmented and tokenized sequences of words. In order to address the rare word problem that abounds in languages with complex morphology, we include a character-based word representation that uses an LSTM to produce embeddings from sequences of characters. Our system was ranked first according to all five relevant metrics for the system: UPOS tagging (93.09%), XPOS tagging (82.27%), unlabeled attachment score (81.30%), labeled attachment score (76.30%), and content word labeled attachment score (72.57%).

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

DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs (2019, methodology): “We used the AllenNLP implementations of state-of-theart models for all of these representations (Gardner et al., 2017; Dozat et al., 2017; He et al., 2017; Stanovsky et al., 2018).”
An Analysis of Encoder Representations in Transformer-Based Machine Translation (2018, methodology): “For comparison purposes, in this dataset, a state of the art supervised parser (Dozat et al., 2017) reaches 88.”
Are Red Roses Red? Evaluating Consistency of Question-Answering Models (2019, methodology): “To generate such instances, we use a dependency parser (Dozat et al., 2017) to recognize root/subject/object and build the implication appropriately, and to detect auxiliary/copula that may need to be moved.”
A Monolingual Approach to Contextualized Word Embeddings for Mid-Resource Languages (2020, methodology): “The parser-speciﬁc bidirectional LSTM layer, whose output is then passed to a bi-afﬁne attention layer (Dozat and Manning, 2017) producing labeled dependency trees.”
Transforming Question Answering Datasets Into Natural Language Inference Datasets (2018, methodology): “We used the state-of-the-art Stanford Graph-Based Neural De-pendency Parser (Dozat et al., 2017) to POS tag and parse Q and A .”
From POS tagging to dependency parsing for biomedical event extraction (2018, methodology): “For Stanford-Biaffine, we use default hyper-parameter values [30].”
Robust Multilingual Part-of-Speech Tagging via Adversarial Training (2017, methodology): “Dozat et al. (2017) empirically show that their dependency parser gains signiﬁcant improvements by using POS tags predicted by a Bi-LSTM POS tagger, while POS tags predicted by the UDPipe tagger (Straka et al., 2016) do not contribute to parsing performance as much.”
Twitter Universal Dependency Parsing for African-American and Mainstream American English (2018, background): “Second, the Deep Biafﬁne ARK Tagger settings produced larger gaps (8.0–11.6”
Cross-lingual Transfer Learning for Multilingual Task Oriented Dialog (2018, methodology): “Here, we follow Dozat et al. (2017) in having a fixed pre-trained embedding matrix combined with tuneable zero-embeddings.”
Multilingual is not enough: BERT for Finnish (2019, methodology): “The dependency parser used in the HIT-SCIR system is the biafﬁne graph-based parser of Dozat et al. (2017) with deep contextualized word embeddings (ELMo) (Peters et al., 2018a) trained monolin-gually on web crawl and Wikipedia data provided by Ginter et al. (2017).”
A Challenge Set and Methods for Noun-Verb Ambiguity (2018, methodology): “For Dozat et al. (2017) and Bohnet et al. (2018), we trained the model on WSJ PTB training data to get comparable models to the two previous systems.”