Papers by Carlos Gómez-Rodríguez

Transition-based Semantic Dependency Parsing with Pointer Networks (2020.acl-main)

Challenge:	Existing work on dependency parsing by sequence labeling suggested that it was impractical.
Approach:	They propose to use dependency trees as sequence labels to obtain fast and accurate parsers using a conventional BILSTM-based model.
Outcome:	The proposed models are conceptually simple, not needing traditional parsing algorithms or auxiliary structures, and provide a good speed-accuracy tradeoff, with results competitive with more complex approaches.

Copied to clipboard

Parsing linearizations appreciate PoS tags - but some are fussy about errors (2022.aacl-short)

Challenge:	Existing dependency parsers cannot be directly applied, so they need to be adaptable to deal with the absence of singlehead and connectedness constraints.
Approach:	They propose a transition system that produces labelled directed acyclic graphs and performs semantic dependency parsing with Pointer Networks.
Outcome:	The proposed system outperforms graph-based models and outperformed existing models on a harder NLP problem.

Copied to clipboard

Alberto Muñoz-Ortiz, Mark Anderson, David Vilares, Carlos Gómez-Rodríguez

Improving Coverage and Runtime Complexity for Exact Inference in Non-Projective Transition-Based Dependency Parsers (N18-2)

Challenge:	Recent work on the impact of PoS tags on graph- and transition-based parsers suggests that they are only useful when tagging accuracy is prohibitively high or in low-resource scenarios.
Approach:	They examine the impact of PoS tags on graph- and transition-based parsers and propose to use them in a new paradigm for sequence labeling.
Outcome:	The proposed model is best when tagging accuracy and resource availability are high.

Copied to clipboard

Tianze Shi, Carlos Gómez-Rodríguez, Lillian Lee

Left-to-Right Dependency Parsing with Pointer Networks (N19-1)

Challenge:	Non-projective dependency trees account for 12.59% of all training sentences in the annotated Universal Dependencies (UD) 2.1 data.
Approach:	They generalize Cohen et al.'s (2011) parser to a family of non-projective transition-based dependency parsers allowing polynomial-time exact inference.
Outcome:	The proposed system can be extended to include a variant that reduces time complexity to O(n6), improving over the known bounds in exact inference for non-projective transition-based parsing.

Copied to clipboard

More Aligned, Less Diverse? Analyzing the Grammar and Lexicon of Two Generations of LLMs (2026.acl-long)

Challenge:	a new algorithm that parses sentences from left to right is simpler than the top-down stack-pointer parser . a graph-based dependency parsing model has been ahead of the curve in terms of accuracy in the past two years .
Approach:	They propose a transition-based algorithm that parses sentences from left to right by building n attachments, with n being the length of the input sentence.
Outcome:	The proposed algorithm outperforms the top-down stack-pointer parser and is twice as fast as the original top-up stack-pointing parsers.

Copied to clipboard

Adrián Gude, Roi Santos-Rios, Francis Bond, Dan Flickinger, Carlos Gómez-Rodríguez, Olga Zamaraeva

Constituent Parsing as Sequence Labeling (D18-1)

Challenge:	a growing number of studies compare LLMs with human-authored text . diversity is unclear, but it is important to understand what makes human and machine writing distinct .
Approach:	They compare syntactic properties of AI-generated and human-authored English news texts . they use the Head-Driven Phrase Structure Grammar and the English Resource Grammar .
Outcome:	The proposed model differs from human-authored English news text in two years.

Copied to clipboard

Carlos Gómez-Rodríguez, David Vilares

Challenge:	Constituent parsing is a core problem in NLP where the goal is to obtain the syntactic structure of sentences expressed as a phrase structure tree.
Approach:	They propose a method to reduce constituent parsing to sequence labeling by using a tree with unary branches.
Outcome:	The proposed method outperforms the Vinyals et al. (2015) sequence-to-sequence parser by 90% on the PTB and CTB treebanks.

Comparing LLM-generated and human-authored news text using formal syntactic theory (2025.acl-long)

Copied to clipboard

Olga Zamaraeva, Dan Flickinger, Francis Bond, Carlos Gómez-Rodríguez

Dependency Graph Parsing as Sequence Labeling (2024.emnlp-main)

Challenge:	a systematic comparison of LLM-generated and human-authored texts is a topic of growing interest in the field of natural language processing.
Approach:	They compare LLM-generated and human-authored New York Times texts using a formal syntactic theory . they use a broad-coverage English resource grammar to analyze the texts .
Outcome:	The proposed comparisons reveal systematic differences between human and LLM-generated texts . the authors hope the results will lead to further discoveries about grammatical properties of LLMs .

Copied to clipboard

Ana Ezquerro, David Vilares, Carlos Gómez-Rodríguez

A Dynamic Oracle for Linear-Time 2-Planar Dependency Parsing (N18-2)

Challenge:	Various linearizations have been proposed to cast syntactic dependency parsing as sequence labeling, but they cannot handle reentrancy or cycles.
Approach:	They propose unbounded linearizations that can be used to cast dependency parsing as sequence labeling.
Outcome:	The proposed linearizations can cast syntactic dependency parsing as a sequence labeling task.

Copied to clipboard

Global Transition-based Non-projective Dependency Parsing (P18-1)

Challenge:	Existing dynamic oracles for greedy parsers can handle non-projective syntax, but none are available for these types of training.
Approach:	They propose an efficient dynamic oracle for training the 2-Planar transition-based parser with over 99% coverage on non-projective syntactic corpora.
Outcome:	The proposed model outperforms the static training strategy in the vast majority of languages tested and scored better on most datasets than the arc-hybrid parser enhanced with the Swap transition.

Copied to clipboard

Carlos Gómez-Rodríguez, Tianze Shi, Lillian Lee

Data Augmentation via Subtree Swapping for Dependency Parsing of Low-Resource Languages (2020.coling-main)

Challenge:	Until recently, transition-based dependency parsers were limited to approximate inference due to their incompatibility with rich feature models.
Approach:	They propose a transition-based parser with high coverage on non-projective treebanks to support non- projective parsing.
Outcome:	The proposed approach is more efficient than its projective counterpart in non-projective languages.

Copied to clipboard

Mathieu Dehouck, Carlos Gómez-Rodríguez

Responsible & Trustworthy NLP

Challenge:	Lack of annotated training data is a big issue for building reliable NLP systems for most of the world’s languages.
Approach:	They propose a method to swap subtrees between annotated sentences while enforcing strong constraints on those trees to ensure maximum grammaticality of the new sentences.
Outcome:	The proposed method outperforms previous methods using the same inputs and using low-resource languages.

Bracketing Encodings for 2-Planar Dependency Parsing (2020.coling-main)

Copied to clipboard

Michalina Strzyz, David Vilares, Carlos Gómez-Rodríguez

HEAD-QA: A Healthcare Dataset for Complex Reasoning (P19-1)

Challenge:	Existing bracketing-based encodings only handle a mild extension of projective trees . encodes that encode arcs in a given plane provide almost total coverage of crossing arc .
Approach:	They propose a bracketing-based encoding that can be used to represent any 2-planar dependency tree over a sentence of length n as a sequence of n labels.
Outcome:	The proposed method improves over existing bracketing encodings in non-projective treebanks while achieving similar speed.

Copied to clipboard

4 and 7-bit Labeling for Projective and Non-Projective Dependency Trees (2023.emnlp-main)

Reasoning

Challenge:	Recent progress in question answering has been led by neural models, but current methods are too data intensive and weak.
Approach:	They propose a multi-choice question answering testbed to encourage research on complex reasoning.
Outcome:	The proposed dataset is useful as a benchmark for future work.

Copied to clipboard

Carlos Gómez-Rodríguez, Diego Roca, David Vilares

Towards Making a Dependency Parser See (D19-1)

Challenge:	encodings that map trees into sequences of one discrete label per word have been proposed for constituency parsing and dependency parse.
Approach:	They propose an encoding that can represent any projective dependency tree as a sequence of 4-bit labels, one per word.
Outcome:	The proposed encoding achieves substantial accuracy gains over the previously best-performing sequence labeling encoders.

Copied to clipboard

Michalina Strzyz, David Vilares, Carlos Gómez-Rodríguez

Spanish Resource Grammar Version 2023 (2024.lrec-main)

Challenge:	Eye trackers and gaze features collected from them have been recently applied to natural language processing (NLP) tasks such as part-of-speech tagging.
Approach:	They propose to leverage eye-tracking data in an RNN dependency parser when no aggregated or token-level gaze features are used at inference time.
Outcome:	The proposed model can be used to improve performance on non-gazed treebanks.

Copied to clipboard

Olga Zamaraeva, Lorena S. Allegue, Carlos Gómez-Rodríguez

Linguistic Theories

Challenge:	Using the Freeling morphological analyzer, we encode a strict notion of grammaticality in the Spanish resource grammar.
Approach:	They propose to use the HPSG formalism to encode a Spanish resource grammar with a manually verified treebank of 2,291 sentences.
Outcome:	The proposed grammars encode a complex set of hypotheses about syntax and a strict notion of grammaticality making them a resource for natural language processing applications in computer-assisted language learning.

A Transition-Based Algorithm for Unrestricted AMR Parsing (N18-2)

Copied to clipboard

Harry Potter and the Action Prediction Challenge from Natural Language (N19-1)

Semantic Text Processing

Challenge:	Abstract Meaning Representation (AMR) is a semantic representation language to map the meaning of English sentences into directed, cycled, labeled graphs.
Approach:	They propose a left-to-right non-projective transition-based parser that handles cycles and reentrancy natively within the transition system itself.
Outcome:	The proposed algorithm handles reentrancy and arbitrary cycles natively, i.e. within the transition system itself.

Copied to clipboard

Discontinuous Constituent Parsing as Sequence Labeling (2020.emnlp-main)

Textual Inference

Challenge:	Using textual descriptions of scenes, we explore the challenge of action prediction from textual description.
Approach:	They propose a testbed to approximate whether text inference can be used to predict upcoming actions from textual descriptions of scenes.
Outcome:	The proposed model performs best for frequent actions and large scene descriptions, but logistic regression fails on infrequent actions.

Copied to clipboard

Revisiting Supertagging for faster HPSG parsing (2024.emnlp-main)

Challenge:	Existing approaches to discontinuous parsing are complex and low-level.
Approach:	They propose to encode discontinuities as nearly ordered permutations of the input sequence.
Outcome:	The proposed model is fast and accurate under the right representation.

Copied to clipboard

Olga Zamaraeva, Carlos Gómez-Rodríguez

Bringing Emerging Architectures to Sequence Labeling in NLP (2026.eacl-long)

Challenge:	a new supertagger for HPSG-based treebanks is used to improve parsing speed and accuracy.
Approach:	They propose to integrate the best supertagger into an HPSG-based parser and compare it to an existing system.
Outcome:	The proposed system achieves 97.26% accuracy on 950 sentences from WSJ23 and 93.88% on the out-of-domain technical essay The Cathedral and the Bazaar.

Copied to clipboard

Ana Ezquerro, Carlos Gómez-Rodríguez, David Vilares

Challenge:	Pretrained Transformer encoders are the dominant approach to sequence labeling . however, few have been applied to sequence labels on flat or simplified tasks .
Approach:	They propose to use pretrained Transformer encoders to model relations across words . they find that the architectures adapt well across tagging tasks that vary in complexity .
Outcome:	The proposed architectures perform well across tagging tasks across languages and datasets.

The Fragility of Multi-Treebank Parsing Evaluation (2022.coling-1)

Copied to clipboard

Iago Alonso-Alonso, David Vilares, Carlos Gómez-Rodríguez

Enriched In-Order Linearization for Faster Sequence-to-Sequence Constituent Parsing (2020.acl-main)

Challenge:	a limitation in NLP evaluation lies in the association between solving a dataset versus solving . authors often run their models only in a handful of treebanks .
Approach:	They propose to run a large-scale experiment on a single treebank and compare them on many parsers whose scores are available.
Outcome:	The proposed model can be biased on a single treebank and spurious effects can be avoided.

Copied to clipboard

Non-Projective Dependency Parsing with Non-Local Transitions (N18-2)

Language Models

Challenge:	Sequence-to-sequence constituent parsing requires a linearization to represent trees as sequences. Top-down tree linearizations have achieved the best accuracy to date.
Approach:	They propose to use an in-order shift-reduce linearization instead of a top-down tree linearization to represent trees as sequences.
Outcome:	The proposed approach achieves the best accuracy to date on the English PTB dataset among fully-supervised single-model sequence-to-sequence constituent parsers.

Copied to clipboard

Dynamic Oracles for Top-Down and In-Order Shift-Reduce Constituent Parsing (D18-1)

Challenge:	Existing greedy transition-based parsers are prone to error propagation when creating arcs involving multiple transitions.
Approach:	They propose a greedy transition-based parser that introduces non-local transitions that create arcs involving nodes to the left of the current focus positions.
Outcome:	The proposed system outperforms the original version and achieves the best accuracy on the Stanford Dependencies conversion of the Penn Treebank among greedy transition-based parsers.

Copied to clipboard

A Confederacy of Models: a Comprehensive Evaluation of LLMs on Creative Writing (2023.findings-emnlp)

Challenge:	Top-down and in-order shift-reduce constituent parsers are the most accurate known shift-reducing algorithms for constituent paring.
Approach:	They propose to use dynamic oracles to train two of the most accurate shift-reduce algorithms for constituent parsing.
Outcome:	The proposed top-down and in-order shift-reduce parsers improve on the WSJ benchmark.

Copied to clipboard

Carlos Gómez-Rodríguez, Paul Williams

Reasoning

Challenge:	a range of recent LLMs on English creative writing have been evaluated . we examine the implications and limitations of our study and suggest directions for future research.
Approach:	They evaluate a range of recent LLMs on English creative writing . they ask LLM writers and humans to write a story using an open-ended scenario .
Outcome:	The results show that state-of-the-art commercial LLMs outperform writers in most dimensions . human creativity is retained, while humor shows a binary divide between LLM that can handle it .

From Partial to Strictly Incremental Constituent Parsing (2024.eacl-short)

Copied to clipboard

Ana Ezquerro, Carlos Gómez-Rodríguez, David Vilares

Challenge:	Incremental NLP aims to learn and adapt partial representations as information unfolds, but studies on incremental approaches have focused on non-incremental approaches.
Approach:	They build incremental constituent parsers that adhere to a strong definition of incrementality across languages and compare them against non-incremental and partially incremental models.
Outcome:	The proposed models are compared against non-incremental and partially incremental models.

Replicating and Extending “Because Their Treebanks Leak”: Graph Isomorphism, Covariants, and Parser Performance (2021.acl-short)

Copied to clipboard

Mark Anderson, Anders Søgaard, Carlos Gómez-Rodríguez

Sequence Labeling Parsing by Learning across Representations (P19-1)

Challenge:	a small sample size and unreliable results suggest a correlation between parser performance and graph isomorphism is not observed in the wild.
Approach:	They propose to replicate a study which found graph isomorphism is a non-trivial variable . they also bin sentences by length and find correlation between parser performance and isopathism disappears .
Outcome:	The results show that the original analysis was unreliable and had methodological issues . the study also bin sentences by length and shows that the correlation between parser performance and graph isomorphism disappears when controlling for covariants.

Copied to clipboard

Michalina Strzyz, David Vilares, Carlos Gómez-Rodríguez

Representation Learning

Challenge:	Constituency and dependency parsing are the main abstractions for representing syntactic structure of sentences . constituency parsers are considered disjointed tasks, and their improvements have been obtained separately.
Approach:	They propose to add auxiliary loss to constituency parsing paradigms and explore a model that parses both paradigms at no cost.
Outcome:	The proposed model outperforms single-task models by 1.05 F1 points and 0.62 UAS points for constituency parsing and dependency parsers.

Hierarchical Bracketing Encodings Work for Dependency Graphs (2025.emnlp-main)

Copied to clipboard

Ana Ezquerro, Carlos Gómez-Rodríguez, David Vilares

Structured Data in NLP

Challenge:	Sequence labeling (SL) is a simple yet effective paradigm for a wide range of natural language problems.
Approach:	They propose a new bracketing approach for dependency graph parsing that encodes graphs as sequences and n tagging actions.
Outcome:	The proposed approach significantly reduces label space while preserving structural information.

Hierarchical Bracketing Encodings for Dependency Parsing as Tagging (2025.acl-long)

Copied to clipboard

Ana Ezquerro, David Vilares, Anssi Yli-Jyrä, Carlos Gómez-Rodríguez

Cross-Lingual Word Embeddings for Turkic Languages (2020.lrec-1)

Challenge:	Existing encodings for dependency parsing use suboptimal number of labels and a limited number of symbols.
Approach:	They propose a family of encodings for sequence labeling dependency parsing based on hierarchical bracketing . they propose an optimal hierarchically bracketing which minimizes the number of symbols used and encodes projective trees using only 12 distinct labels .
Outcome:	The proposed encodings yield competitive accuracy on a diverse set of treebanks.

Copied to clipboard

Elmurod Kuriyozov, Yerai Doval, Carlos Gómez-Rodríguez

Multilinguality

Challenge:	Existing techniques to align monolingual embeddings are difficult to use because of low resources.
Approach:	They propose to use existing techniques to align monolingual embedding spaces for Turkic, Uzbek, Azeri, Kazakh and Kyrgyz languages.
Outcome:	The proposed techniques outperform existing techniques on bilingual dictionaries and an extrinsic task.

Inherent Dependency Displacement Bias of Transition-Based Algorithms (2020.lrec-1)

Copied to clipboard

Mark Anderson, Carlos Gómez-Rodríguez

Nested Named Entity Recognition as Single-Pass Sequence Labeling (2025.findings-emnlp)

Challenge:	Empirical studies have shown that performance varies across different treebanks in such a way that one algorithm outperforms another on one treebank and the reverse is true for a different tree bank.
Approach:	They introduce the concept of an algorithm’s inherent dependency displacement distribution and characterise its bias in terms of dependency displacement.
Outcome:	The proposed model shows that the similarity of an algorithm’s inherent dependency displacement distribution to a treebank’s displacement distribution is clearly correlated to the algorithm’ s parsing performance on that treebank.

Copied to clipboard

Alberto Muñoz-Ortiz, David Vilares, Caio Corro, Carlos Gómez-Rodríguez

Information Extraction & Text Mining

Challenge:	nested named entity recognition is a sequence labeling task that can be trained using any off-the-shelf sequence labelling library.
Approach:	They use prior work that linearizes constituency structures to create a nested named entity recognition task.
Outcome:	The proposed method reduces the complexity of the predicted nested entity recognition problem to a simple token classification task.

Reducing Discontinuous to Continuous Parsing with Pointer Network Reordering (2021.emnlp-main)

Copied to clipboard