Papers by Trevor Cohn
Fairness-aware Class Imbalanced Learning (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on class imbalance and mitigating bias have focused on the latter . a skewed class distribution hurts the performance of deep learning models, and is often referred to as "stereotyping" |
| Approach: | They propose to extend a margin-loss based approach to enforce fairness by using tweet sentiment and occupation classification to mitigate class imbalance and demographic bias. |
| Outcome: | The proposed methods help mitigate class imbalance and demographic biases through controlled experiments. |
FairLib: A Unified Framework for Assessing and Improving Fairness (2022.emnlp-demos)
Copied to clipboard
| Challenge: | Existing approaches to assess and improve model fairness have been inconsistent and inconsistent. |
| Approach: | They propose an open-source python library for assessing and improving model fairness. |
| Outcome: | The proposed framework can be used for natural language, images, and audio. |
Cost-effective Distillation of Large Language Models (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing knowledge distillation methods require pretraining of the teacher on task-specific datasets, which can be costly for large and unstable for small datasets. |
| Approach: | They propose an approach to improve knowledge distillation by a loss-agnostic approach to task and model architecture. |
| Outcome: | The proposed method achieves competitive results across a range of tasks, especially for tasks with smaller datasets. |
Evaluating Debiasing Techniques for Intersectional Biases (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for debiasing protected attributes have been limited to binary attributes in isolation, however many corpora involve multiple such attributes, possibly with higher cardinality. |
| Approach: | They propose to evaluate a bias-constrained model which is new to NLP and an extension of the iterative nullspace projection technique which can handle multiple identities. |
| Outcome: | The proposed model is based on a new iterative nullspace projection technique which can handle multiple identities. |
Narrative Modeling with Memory Chains and Semantic Supervision (P18-2)
Copied to clipboard
| Challenge: | Story comprehension requires a deep semantic understanding of the narrative, making it a challenging task. |
| Approach: | They propose a method that tracks various semantic aspects with external neural memory chains . they propose to encourage each to focus on a particular semantic aspect . |
| Outcome: | The proposed method outperforms baselines on the task of story ending prediction. |
Incorporating Syntax and Semantics in Coreference Resolution with Heterogeneous Graph Attention Network (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing neural coreference resolution models lack syntactic and semantic information . however, such information has been shown to benefit other tasks. |
| Approach: | They propose a graph-based model that incorporates syntactic and semantic structures of sentences. |
| Outcome: | The proposed model incorporates syntactic and semantic structures of sentences. |
It Is Not As Good As You Think! Evaluating Simultaneous Machine Translation on Interpretation Data (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing siMT systems are trained and evaluated on offline translations . however, evaluation gap remains notable, calling for constructing large-scale interpretation corpora . |
| Approach: | They propose a translation-to-interpretation transfer method which converts offline translations into interpretation-style data. |
| Outcome: | The proposed interpretation test set shows that SiMT models improve on translation vs interpretation data. |
Neural Speech Translation using Lattice Transformations and Graph Networks (D19-53)
Copied to clipboard
| Challenge: | Existing work on end-to-end systems bypass the need for intermediate representations, but this approach is limited in practical applications. |
| Approach: | They propose a lattice-tosequence model which uses lattics as encoders and graph networks to address two problems by applying latticae transformations and a neural model. |
| Outcome: | The proposed model beats pipeline approaches while being orders of magnitude faster than previous work. |
Semi-supervised User Geolocation via Graph Convolutional Networks (P18-1)
Copied to clipboard
| Challenge: | Social media user geolocation is vital to many applications such as event detection. |
| Approach: | They propose a multiview geolocation model that uses both text and network context. |
| Outcome: | The proposed model outperforms baseline models and the state-of-the-art models under minimal supervision. |
Optimising Equal Opportunity Fairness in Model Training (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing methods to reduce bias have been shown to be effective over real-world datasets. |
| Approach: | They propose two new training objectives which directly optimise for the widely-used criterion of equal opportunity. |
| Outcome: | The proposed training objectives directly optimise for the widely-used criterion of equal opportunity while maintaining high performance over two classification tasks. |
Simpson’s Paradox and the Accuracy-Fluency Tradeoff in Translation (2024.acl-short)
Copied to clipboard
| Challenge: | Existing studies suggest that accuracy and fluency should trade off against each other, and that capturing every detail of the source is difficult for human raters to distinguish. |
| Approach: | They propose to evaluate the relationship between accuracy and fluency at the segment level and to use probabilities to estimate probabilities. |
| Outcome: | The proposed model relies on human judgments of accuracy and fluency collected in prior work on translation quality estimation. |
Mitigating Backdoor Poisoning Attacks through the Lens of Spurious Correlation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Modern NLP models are often trained over large untrustworthy datasets, raising the potential for a malicious adversary to compromise model behaviour. |
| Approach: | They propose to mitigate spurious correlations between textual triggers and classification labels by combining them with insertion-based attacks. |
| Outcome: | The proposed defence significantly reduces attack success rates across backdoor attacks and provides a near-perfect defence against insertion-based attacks. |
Predicting Human Translation Difficulty Using Automatic Word Alignment (2023.findings-acl)
Copied to clipboard
| Challenge: | Translation difficulty is a problem when translators are required to resolve translation ambiguity from multiple possible translations. |
| Approach: | They use word alignments computed over large scale bilingual corpora to develop predictors of lexical translation difficulty. |
| Outcome: | The proposed method improves on a previous embedding-based approach and can contribute to a deeper understanding of cross-lingual differences and of causes of translation difficulty. |
Tokenizer-Aware Cross-Lingual Adaptation of Decoder-Only LLMs through Embedding Relearning and Swapping (2026.eacl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have been primarily focused on English, leaving the multilingual ability unexplored. |
| Approach: | They propose a technique that creates new tokenizers and tunes embeddings on fixed model weights for target language adaptation. |
| Outcome: | The proposed method is light-weight and performant but has limitations for older models and high resource languages. |
A Stochastic Decoder for Neural Machine Translation (P18-1)
Copied to clipboard
| Challenge: | Neural machine translation models do not account for local lexical and syntactic variation in parallel corpora. |
| Approach: | They propose a deep generative model of machine translation which incorporates a chain of latent variables to account for local lexical and syntactic variation in parallel corpora. |
| Outcome: | The proposed model consistently improves over strong baselines on several different language pairs. |
Probing Power by Prompting: Harnessing Pre-trained Language Models for Power Connotation Framing (2023.eacl-main)
Copied to clipboard
| Challenge: | Using pre-trained language models, we investigated whether word choices can encode subtle connotative information about power differentials between involved entities. |
| Approach: | They propose a framework to disentangle connotation frames implied by the predicate from its arguments and the sentence structure and to quantify predicates. |
| Outcome: | The proposed framework improves power connotation prediction accuracy by fine-tuning pre-trained language models. |
Evaluating the Utility of Hand-crafted Features in Sequence Labelling (D18-1)
Copied to clipboard
| Challenge: | Conventional wisdom is that hand-crafted features are redundant for deep learning models . authors propose a method for using handcrafted features in a hybrid learning approach . |
| Approach: | They propose a method for exploiting handcrafted features as part of a hybrid learning approach. |
| Outcome: | The proposed method outperforms baseline models on a named entity recognition task and reduces training requirements to 60% while maintaining the same predictive accuracy. |
Measuring and Mitigating Name Biases in Neural Machine Translation (2022.acl-long)
Copied to clipboard
| Challenge: | Neural machine translation systems exhibit problematic biases, such as stereotypical gender bias in occupation terms. |
| Approach: | They propose a method to reduce biases in person name translations by randomly switching entities during translation. |
| Outcome: | The proposed method eliminates the problem without any effect on translation quality. |
Unsupervised Cross-Lingual Transfer of Structured Predictors without Source Data (2022.naacl-main)
Copied to clipboard
| Challenge: | Recent successes of NLP systems require large amounts of labelled data for structured prediction tasks. |
| Approach: | They propose a method for unsupervised transfer from multiple input models for structured prediction using a cross-lingual setup. |
| Outcome: | The proposed method produces less noisy labels for the distant supervision. |
Massively Multilingual Transfer for NER (P19-1)
Copied to clipboard
| Challenge: | Existing approaches for cross-lingual transfer use a single source language, but there are exceptions. |
| Approach: | They propose two techniques for modulating the transfer, suitable for zero-shot or few-shot learning, respectively. |
| Outcome: | The proposed methods are much more effective than baseline models and rival oracle selection of the single best individual model. |
Tangled up in BLEU: Reevaluating the Evaluation of Automatic Machine Translation Evaluation Metrics (2020.acl-main)
Copied to clipboard
| Challenge: | Existing methods for judging metrics are sensitive to the translations used for evaluation, leading to falsely confident conclusions about a metric’s efficacy. |
| Approach: | They propose a method for thresholding performance improvement under an automatic metric against human judgements by using a pairwise system ranking method. |
| Outcome: | The proposed method allows quantification of type I versus type II errors incurred, i.e., insignificant human differences in system quality that are accepted, and significant human differences that are rejected. |
WAX: A New Dataset for Word Association eXplanations (2022.aacl-main)
Copied to clipboard
| Challenge: | Word associations are among the most common paradigms to study the human mental lexicon. |
| Approach: | They present a large dataset of word associations with explanations and relation labels . they show that current language models struggle to capture the diversity of human associations . |
| Outcome: | The proposed model fails to capture the diversity of human associations, the authors show . they show that the model is a rich benchmark for commonsense modeling and generation. |
DeltaScore: Fine-Grained Story Evaluation with Perturbations (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing evaluation metrics for stories are limited in assessing intricate aspects of storytelling, such as fluency and interestingness. |
| Approach: | They propose a novel method that uses perturbation techniques to evaluate story aspects . they compare fluency, coherence, relatedness, logicality, interestingness and interestingness to existing metrics . |
| Outcome: | The proposed method shows that one specific perturbation is highly effective in capturing multiple aspects. |
Hierarchical Structured Model for Fine-to-Coarse Manifesto Text Analysis (N18-1)
Copied to clipboard
| Challenge: | Election manifestos document the intentions, motives, and views of political parties. |
| Approach: | They propose a hierarchical structured deep model to predict fine- and coarse-grained positions and a probabilistic soft logic model to perform post-hoc calibration of coarse- and fine-grain positions. |
| Outcome: | The proposed model outperforms state-of-the-art approaches at both granularities using manifestos from twelve countries, written in ten different languages. |
Recurrent Entity Networks with Delayed Memory Update for Targeted Aspect-Based Sentiment Analysis (N18-2)
Copied to clipboard
| Challenge: | Recent work on target-dependent biLSTMs has shown that they are ineffective in aspect-based sentiment analysis. |
| Approach: | They propose a novel architecture that uses external memory chains with a delayed memory update mechanism to track entities. |
| Outcome: | The proposed model improves on a TABSA task using external memory chains with a delayed memory update mechanism. |
Balancing out Bias: Achieving Fairness Through Balanced Training (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to reducing group bias do not account for correlations between author demographics and linguistic variables, limiting their effectiveness. |
| Approach: | They extend a method for countering group bias using balanced training by balancing each demographic group in training and using protected attributes as input. |
| Outcome: | The proposed model outperforms all other methods when combined with balanced training. |
Not another Negation Benchmark: The NaN-NLI Test Suite for Sub-clausal Negation (2022.aacl-main)
Copied to clipboard
| Challenge: | Negation is an important linguistic phenomenon which denotes non-existence, denial, or contradiction. |
| Approach: | They propose a natural language inference test suite to test models for negation . they use a linguistic framework to analyze negation types and constructions . |
| Outcome: | The proposed test suite is more challenging than existing benchmarks on negation . it includes annotation of negation types and constructions grounded in linguistic theory . |
Putting Evaluation in Context: Contextual Embeddings Improve Machine Translation Evaluation (P19-1)
Copied to clipboard
| Challenge: | Existing evaluation metrics are limited and can be easily portable to new languages. |
| Approach: | They propose a simple unsupervised metric and additional supervised metrics which rely on contextual word embeddings to encode the translation and reference sentences. |
| Outcome: | The proposed model outperforms existing metrics on the WMT 2017 dataset and is more accurate than existing models. |
What’s in a Domain? Learning Domain-Robust Text Representations using Adversarial Training (N18-2)
Copied to clipboard
| Challenge: | a key roadblock is application to new domains, unseen in training. |
| Approach: | They propose a method to optimise in- and out-of-domain accuracy by combing domain-specific and domain-general components with adversarial training for domain. |
| Outcome: | The proposed method improves on domain adaptation and domain-adversarial training. |
Systematic Evaluation of Predictive Fairness (2022.aacl-main)
Copied to clipboard
| Challenge: | Several methods have been proposed to mitigate bias in training on biased datasets. |
| Approach: | They propose to examine the effect of target class imbalance and stereotyping on model performance by analyzing binary classification, profession prediction and regression tasks. |
| Outcome: | The proposed methods show that data conditions have a strong influence on relative model performance. |
Grounding learning of modifier dynamics: An application to color naming (D19-1)
Copied to clipboard
| Challenge: | Existing models for grounding are unable to understand modified color expressions, such as “light blue”. |
| Approach: | They propose a model that learns more complex transformations in RGB space and a hard ensemble model that selects a color space depending on the modifier-color pair. |
| Outcome: | The proposed model performs better in the HSV color space than the state-of-the-art model. |
Evaluation Phonemic Transcription of Low-Resource Tonal Languages for Language Documentation (L18-1)
Copied to clipboard
| Challenge: | Language documentation involves recording the speech of native speakers. |
| Approach: | They propose to use a neural network architecture to model phonemes and tones versus modelling them separately. |
| Outcome: | The proposed method improves efficiency, minimizes typographical errors and maintains transcription faithfulness to acoustic signal while highlighting phonetic and phonemic facts for linguistic consideration. |
Putting words into the system’s mouth: A targeted attack on neural machine translation using monolingual data poisoning (2021.findings-acl)
Copied to clipboard
Jun Wang, Chang Xu, Francisco Guzmán, Ahmed El-Kishky, Yuqing Tang, Benjamin Rubinstein, Trevor Cohn
| Challenge: | Neural machine translation systems are known to be vulnerable to adversarial test inputs, however, they are also vulnerable to training attacks. |
| Approach: | They propose a poisoning attack in which a malicious adversary inserts a small poisoned sample of monolingual text into a training set of a system trained using back-translation. |
| Outcome: | The proposed attack is based on two methods that can be used to craft poisoned examples. |
Decoupling Adversarial Training for Fair NLP (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing work assumes main task labels and protected attributes are available in the dataset, but protected labels are often unavailable or only available in limited numbers. |
| Approach: | They propose a method which uses only a small volume of protected labels to train adversarial models using a dataset with a discriminator. |
| Outcome: | The proposed method can be used to transfer private-labelled instances from one dataset to another without requiring large amounts of protected labels. |
Deep Ordinal Regression for Pledge Specificity Prediction (D19-1)
Copied to clipboard
| Challenge: | Currently, there are no publicly available annotated datasets of pledges . a novel approach to specificity prediction is needed to predict the specificity of pledged issues. |
| Approach: | They propose deep ordinal regression approaches for specificity prediction using supervised and semi-supervised settings. |
| Outcome: | The proposed methods demonstrate their utility over several baseline approaches. |
Content-based Popularity Prediction of Online Petitions Using a Deep Regression Model (P18-2)
Copied to clipboard
| Challenge: | Existing work on predicting popularity of online petitions based on initial popularity trajectory has focused on estimating the number of signatures a petition gets in the first x hours, and predicting the total number of signed petitions at the end of its lifetime. |
| Approach: | They propose a CNN-based model to predict the popularity of a petition based on its textual content and use it to model the influence of other petition signers. |
| Outcome: | The proposed model is based on UK and US government petition datasets and is compared with previous work on predicting popularity over time based upon initial popularity trajectory. |
SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing safeguard models rely on translation of English datasets, missing regional and cultural nuances. |
| Approach: | They propose a framework to generate culturally grounded safety datasets for Southeast Asia . SEA-Guard family is the first multilingual safeguard model grounded in SEA cultural contexts . |
| Outcome: | The proposed model outperforms existing safeguard models in detecting regionally sensitive content while maintaining strong general safety performance. |
Rethinking Round-Trip Translation for Machine Translation Evaluation (2023.findings-acl)
Copied to clipboard
| Challenge: | Automatic evaluation methods for translation often require model training and the availability of parallel corpora limits their applicability to low-resource settings. |
| Approach: | They revisit the statistical machine translation technique and use it to improve translation quality. |
| Outcome: | The proposed method improves translation quality estimation models and identifies adversarial competitors in shared tasks via cross-system verification. |
Learning Coupled Policies for Simultaneous Machine Translation using Imitation Learning (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing approaches to learn simultaneous translation model with coupled programmer-interpreter policies are suboptimal as they fix the agent's policy to focus learning the NMT model or learn adaptive agent policies while the NRT model is fixed. |
| Approach: | They propose an algorithmic oracle to produce oracular READ/WRITE actions for training bilingual sentence-pairs using the notion of word alignments. |
| Outcome: | The proposed method outperforms baselines in terms of translation quality quality while keeping the delay low. |
Graph-to-Sequence Learning using Gated Graph Neural Networks (P18-1)
Copied to clipboard
| Challenge: | Existing approaches to graph-to-sequence learning ignore the full graph structure, discarding key information. |
| Approach: | They propose a graph-to-sequence learning model that encodes the full graph structure and an input transformation that allows nodes and edges to have their own hidden representations. |
| Outcome: | The proposed model outperforms baselines in generation from AMR graphs and syntax-based neural machine translation while retaining the full graph structure. |
Noisy Self-Training with Synthetic Queries for Dense Retrieval (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing neural retrieval models require training on a sufficient number of human-labelled query-passage pairs to work well. |
| Approach: | They propose a noisy self-training framework with synthetic queries to improve retrieval methods. |
| Outcome: | The proposed method outperforms baselines on general-domain and out-of-domain retrieval benchmarks on low-resource settings and is data efficient and data efficient. |
Boot and Switch: Alternating Distillation for Zero-Shot Dense Retrieval (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to enhance dense retrieval models are unwieldy, such as requiring explicit supervision, complex model architectures, or massive external models. |
| Approach: | They propose an unsupervised method to enhance passage retrieval in zero-shot settings by iterating a loop that a dense retriever learns from supervision signals provided by a reranker. |
| Outcome: | The proposed method outperforms leading supervised and unsupervised retrievers on the BEIR benchmark while showing strong adaptation abilities to tasks and domains that were unseen during training. |
Revisiting subword tokenization: A case study on affixal negation in large language models (2024.naacl-long)
Copied to clipboard
| Challenge: | Negation is central to language understanding but is not properly captured by modern NLP methods. |
| Approach: | They propose to use subword tokenization methods to detect negation in large language models . they find that models can reliably recognize negation, despite mismatches in tokenization accuracy . |
| Outcome: | The proposed models can detect negation in English using subword tokenization methods despite some mismatches in tokenization accuracy and negation detection performance. |
Diverse Adversaries for Mitigating Bias in Training (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing adversarial methods only partially mitigate the problem of model bias, added to which their training procedures are unstable. |
| Approach: | They propose a method where discriminators are encouraged to learn orthogonal hidden representations from one another to reduce model bias. |
| Outcome: | The proposed method significantly reduces bias and stability of training over standard methods. |
Don’t Mess with Mister-in-Between: Improved Negative Search for Knowledge Graph Completion (2023.eacl-main)
Copied to clipboard
| Challenge: | Existing methods for knowledge graph completion use a dual-encoding framework with a bottleneck that allows for fast approximate search over a vast collection of candidates. |
| Approach: | They propose to use a dual-encoder framework to find more informative negatives by searching for candidates with high lexical overlaps. |
| Outcome: | The proposed methods improve on the large-scale Wikidata5M dataset and combine different kinds of strategies to achieve state-of-the-art performance. |
A Survey for Efficient Open Domain Question Answering (2023.acl-long)
Copied to clipboard
| Challenge: | Open domain question answering (ODQA) is a longstanding task that can answer factoid questions without explicit evidence in natural language processing (NLP). |
| Approach: | They propose to use open domain question answering to answer factual questions from a large knowledge corpus without explicit evidence. |
| Outcome: | The proposed models can answer factoid questions from a large knowledge corpus without explicit evidence. |
Tulun: Transparent and Adaptable Low-resource Machine Translation (2025.acl-demo)
Copied to clipboard
| Challenge: | a low-resource language that is the lingua franca in Timor-Leste lacks available corpora in the health domain. |
| Approach: | They propose a solution that combines neural MT with large language model-based post-editing guided by existing glossaries and translation memories. |
| Outcome: | The proposed system outperforms both standalone MT and LLM approaches across six low-resource languages on the FLORES dataset. |
Performance Prediction via Bayesian Matrix Factorisation for Multilingual Natural Language Processing Tasks (2023.eacl-main)
Copied to clipboard
| Challenge: | Performance prediction for natural language processing (NLP) is based on a framework of Bayesian matrix factorisation . it avoids hyperparameter tuning and provides uncertainty estimates over predictions. |
| Approach: | They propose to use Bayesian matrix factorisation to predict the performance of language pairs depicted by grey cells. |
| Outcome: | The proposed framework outperforms the state-of-the-art in several NLP benchmarks, including machine translation and cross-lingual entity linking. |
Deep-speare: A joint neural model of poetic language, meter and rhyme (P18-1)
Copied to clipboard
| Challenge: | a recent surge of interest in deep learning has led to creative applications for poetry generation . a novel joint architecture captures language, rhyme and meter for sonnet modelling . |
| Approach: | They propose a joint architecture that captures language, rhyme and meter for sonnet modelling. |
| Outcome: | The proposed architecture captures language, rhyme and meter for sonnet modelling. |
Towards Robust and Privacy-preserving Text Representations (P18-2)
Copied to clipboard
| Challenge: | Written text often provides sufficient clues to identify the author, gender, age, and other important attributes. |
| Approach: | They propose to obscure important author characteristics at training time so that models are invariant to these attributes. |
| Outcome: | The proposed approach leads to increased privacy in the learned representations, and robust models to varying evaluation conditions, including out-of-domain corpora. |
Foiling Training-Time Attacks on Neural Machine Translation Systems (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Neural machine translation systems are vulnerable to backdoor attacks . successful backdoors can cause slander, hate speech, phishing, etc. attacks can target very short trigger phrases, which can be challenging to detect even when included verbatim in poisoned instances. |
| Approach: | They propose a method that exploits asymmetry between source and target sentences to detect outlier tokens. |
| Outcome: | The proposed method reduces the success of attacks by up to 89.0% while not affecting predictive accuracy. |
Backdoor Attacks on Multilingual Machine Translation (2024.naacl-long)
Copied to clipboard
| Challenge: | Recent studies have shown that multilingual machine translation systems are vulnerable to backdoor attacks through data poisoning. |
| Approach: | They propose to investigate the security of multilingual machine translation systems by exposing poisoned data into low-resource languages to cause malicious translations. |
| Outcome: | The proposed method achieves an average of 20% success rate in attacking high-resource languages. |
Language models are not naysayers: an analysis of language models on negation benchmarks (2023.starsem-1)
Copied to clipboard
| Challenge: | Negation has been shown to be a major bottleneck for masked language models, such as BERT, but whether this finding still holds for larger-sized auto-regressive language models has not been studied comprehensively. |
| Approach: | They evaluate the ability of current-generation auto-regressive language models to handle negation using a wide range of benchmarks and models. |
| Outcome: | The proposed models are compared against a wide range of negation benchmarks and show that they are insensitive to negation, inability to capture the lexical semantics of negations, and failure to reason under negation. |
Improving negation detection with negation-focused pre-training (2022.naacl-main)
Copied to clipboard
| Challenge: | Negation is a common linguistic feature that is crucial in many language understanding tasks. |
| Approach: | They propose a new approach to detect negation in language models using data augmentation and negation masking. |
| Outcome: | The proposed approach improves negation detection performance and generalizability over the strong baseline NegBERT. |
More than Votes? Voting and Language based Partisanship in the US Supreme Court (2023.findings-emnlp)
Copied to clipboard
| Challenge: | partisanship and ideology have been a key topic in legal studies of the US Supreme Court . most research quantifies partisan behavior based on voting behavior, and oral arguments have not been well studied for this purpose. |
| Approach: | They propose a framework for analyzing justices' oral arguments for partisan signals and how they align with voting patterns. |
| Outcome: | The proposed framework shows that the affiliated party of justices can be predicted reliably from their oral contributions. |
Semi-supervised Stochastic Multi-Domain Learning using Variational Inference (P19-1)
Copied to clipboard
| Challenge: | Supervised NLP models rely on large collections of text which closely resemble intended testing setting. however, data is often messy, with domain labels not always available, or providing limited information about the style and genre of text. |
| Approach: | They propose a method to distill the important domain signal as part of a multi-domain learning system using a latent variable model. |
| Outcome: | The proposed model improves performance over benchmark domain adaptation methods . text corpora are often collated from several different sources, including news, literature, microblogs, and web crawls . |
Mitigating Data Poisoning in Text Classification with Differential Privacy (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Data poisoning attacks can plant a backdoor in a model by injecting poisoned examples into training data, causing the model to misclassify test instances which include a specific pattern. |
| Approach: | They propose a generic defence mechanism that makes training robust to poisoning attacks by smoothing the gradient from each training example. |
| Outcome: | The proposed method is highly effective in mitigating, or even eliminating, poisoning attacks on text classification, with only a small cost in predictive accuracy. |
Does Representational Fairness Imply Empirical Fairness? (2022.findings-aacl)
Copied to clipboard
| Challenge: | Neural methods have been trained on datasets which embody cultural and societal stereotypes, captured in spurious correlations between target labels and protected attributes. |
| Approach: | They propose a debiasing method that encourages a latent space that separates instances based on target label, while mixing instances that share protected attributes. |
| Outcome: | The proposed method shows that representational fairness does not imply empirical fairness across methods. |
Pre-training Cross-lingual Open Domain Question Answering with Large-scale Synthetic Supervision (2024.emnlp-main)
Copied to clipboard
| Challenge: | Cross-lingual open domain question answering requires multiple models, requiring substantial annotated datasets and auxiliary resources to bridge between languages. |
| Approach: | They propose a selfsupervised method that exploits Wikipedia's cross-lingual link structure . they show that the method outperforms comparable methods on supervised and zero-shot settings . |
| Outcome: | The proposed method outperforms comparable methods on supervised and zero-shot language adaptation settings. |
Contextualization of Morphological Inflection (N19-1)
Copied to clipboard
| Challenge: | In this paper, we isolate the task of predicting a fully inflected sentence from its partially lemmatized version. |
| Approach: | They propose a task that requires morphological features to be inferred from sentential context . they propose morphology-based models that explicitly reconstruct morphologic features before predicting inflected forms . |
| Outcome: | The proposed model is able to predict inflected sentences without relying on morphological annotations. |
TUBA: Cross-Lingual Transferability of Backdoor Attacks in LLMs with Instruction Tuning (2025.findings-acl)
Copied to clipboard
Xuanli He, Jun Wang, Qiongkai Xu, Pasquale Minervini, Pontus Stenetorp, Benjamin I. P. Rubinstein, Trevor Cohn
| Challenge: | Despite the increasing support for multilingual capabilities, the impact of backdoor attacks on LLMs remains under-explored. |
| Approach: | They propose to use poisoned instructiontuning data to attack multilingual LLMs . their results show that more powerful models show increased susceptibility to transferable cross-lingual backdoor attacks . |
| Outcome: | The proposed attack is effective in models like BLOOM and GPT-4o with high success rates in more than 7 out of 12 languages. |
PPT: Parsimonious Parser Transfer for Unsupervised Cross-Lingual Adaptation (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing methods for cross-lingual transfer use implicit supervision to parse low-resource languages without explicit supervision. |
| Approach: | They propose a method for unsupervised cross-lingual transfer that uses their output as implicit supervision as part of self-training on unlabelled text in the target language. |
| Outcome: | The proposed method improves over state-of-the-art models on both distant and nearby languages, despite being conceptually simpler. |
Seeking Clozure: Robust Hypernym extraction from BERT with Anchored Prompts (2023.starsem-1)
Copied to clipboard
| Challenge: | Existing methods for extracting hypernym knowledge from large language models are unclear whether they fail due to a lack of knowledge or shortcomings. |
| Approach: | They propose to use pattern-based hypernym extraction as a diagnostic tool to examine hypernomy knowledge encoded in BERT. |
| Outcome: | The proposed method compares the results of two different methods on six English data sets and on challenge sets of rare and abstract concepts. |
On the Role of Scene Graphs in Image Captioning (D19-64)
Copied to clipboard
| Challenge: | Recent captioning approaches rely on ad-hoc approaches to obtain graphs for images, but they introduce noise and it is unclear the effect of parser errors on captioning accuracy. |
| Approach: | They investigate whether scene graphs can help image captioning . they show that a scene graph parser can boost performance almost as much as ground truth graphs . |
| Outcome: | The proposed parser can boost performance almost as much as ground truth graphs . |
Framing Unpacked: A Semi-Supervised Interpretable Multi-View Model of Media Frames (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing models for news analysis lack transparency in their predictions. |
| Approach: | They propose a semi-supervised model that embeds local information into news articles . it can be used to improve automatic news analysis, authors argue . |
| Outcome: | The proposed model outperforms previous models and can be used with unlabeled training data. |
Fair Enough: Standardizing Evaluation and Model Selection for Fairness Research in NLP (2023.eacl-main)
Copied to clipboard
| Challenge: | Modern NLP systems exhibit a range of biases, which a growing literature on model debiasing attempts to correct. |
| Approach: | They propose to clarify the current situation and plot a course for meaningful progress in fair learning by making clear inter-relations among the current gamut of methods and their relation to fairness theory. |
| Outcome: | The proposed approach addresses the practical problem of model selection, which involves a trade-off between fairness and accuracy and has led to systemic issues in fairness research. |
As Easy as 1, 2, 3: Behavioural Testing of NMT Systems for Numerical Translation (2021.findings-acl)
Copied to clipboard
| Challenge: | Mistranslated numbers can cause financial loss or medical misinformation. |
| Approach: | They propose a method to assess the robustness of neural machine translation systems to numerical text via behavioural testing. |
| Outcome: | The proposed method systematically assesses four fundamental capabilities of neural machine translation systems in translation numbers by virtue of a variety of test cases. |
SEA-SafeguardBench: Culturally Grounded Safety Benchmark for Southeast Asian Languages (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing multilingual safety benchmarks rely on machine-translated English data, which fails to capture nuances in low-resource languages. |
| Approach: | They propose to use a human-verified safety benchmark for Southeast Asian languages to validate their safety and cultural diversity. |
| Outcome: | The proposed model outperforms existing models in general, in-the-wild, and content generation across eight languages and 21,640 samples across three subsets: general, and in- the-wild. |
LORAXBENCH: A Multitask, Multilingual Benchmark Suite for 20 Indonesian Languages (2025.emnlp-main)
Copied to clipboard
| Challenge: | LORAXBENCH is a benchmark for low-resource languages of Indonesia . it covers reading comprehension, open domain QA, language inference, causal reasoning, translation, and cultural question answering across 20 languages. |
| Approach: | They propose a benchmark that focuses on low-resource languages of Indonesia and covers 6 diverse tasks: reading comprehension, open-domain QA, language inference, causal reasoning, translation, and cultural question answering. |
| Outcome: | The proposed benchmark covers reading comprehension, open-domain QA, language inference, causal reasoning, translation, and cultural question answering across 20 Indonesian languages. |