Papers by Adam Lopez
A systematic comparison of methods for low-resource dependency parsing on genuinely low-resource languages (D19-1)
Copied to clipboard
| Challenge: | Large annotated treebanks are available for only a tiny fraction of the world's languages, and there is a wealth of literature on strategies for parsing with few resources. |
| Approach: | They propose three strategies for improving low-resource parsers: data augmentation, cross-lingual training, and transliteration. |
| Outcome: | The proposed methods improve low-resource parsers by using data augmentation, cross-lingual training, and transliteration. |
Low-Rank Softmax Can Have Unargmaxable Classes in Theory but Rarely in Practice (2022.acl-long)
Copied to clipboard
| Challenge: | Probabilistic multiclass classifiers with large number of output classes are commonplace in natural language processing. |
| Approach: | They propose to use argmax to predict words from a large vocabulary in NLP models . they find that 13 out of 150 models do indeed have such unargmaxable tokens . |
| Outcome: | The proposed algorithms detect unargmaxable tokens in large language models and translation models. |
Intrinsic Bias Metrics Do Not Correlate with Application Bias (2021.acl-long)
Copied to clipboard
| Challenge: | a recent survey of bias in natural language processing found that a coreference system makes more errors in an anti-stereotypical coreferent than in a pro-sterereotype one. |
| Approach: | They compare intrinsic and extrinsic bias metrics across hundreds of trained models . they urge researchers to focus on extrindic measures of bias, not easy to measure . |
| Outcome: | a new intrinsic metric and an annotated test set on gender bias in hate speech are tested . authors urge researchers to focus on extrinsic measures of bias, and to make them more feasible . |
Inflecting When There’s No Majority: Limitations of Encoder-Decoder Neural Networks as Cognitive Models for German Plurals (2020.acl-main)
Copied to clipboard
| Challenge: | Encoder-decoder models can be used to generalize to inflectional morphology and generalize new words, but they fail on tasks like German number inflection, where infrequent suffixes like /-s/ can still be productively generalized. |
| Approach: | They propose to use a dataset to collect data from German speakers to examine whether ED models can generalize the most frequently produced plural class. |
| Outcome: | The proposed model does not show human-like variability or ‘regular’ extension of other plural markers. |
Understanding Learning Dynamics Of Language Models with SVCCA (N19-1)
Copied to clipboard
| Challenge: | a new study shows that neural models implicitly encode linguistic features . but no research shows how these encodings arise as the models are trained . |
| Approach: | They propose a method that compares learning across time and across models using annotated data. |
| Outcome: | The proposed method compares learned representations across time and across models without evaluation on annotated data. |
LLMs Reproduce Stereotypes of Sexual and Gender Minorities (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a large body of research has found substantial gender bias in NLP systems . authors show that LLMs generate stereotyped representations of sexual and gender minorities in this setting . |
| Approach: | They propose to use a stereotype content model to study gender bias in large language models . they show that LLMs generate stereotyped representations of sexual and gender minorities . |
| Outcome: | The proposed model generates negative stereotypes of sexual and gender minorities in English-language surveys . |
First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | a new system trained on well over a trillion words smashes the state of the art by a margin previously thought impossible. |
| Approach: | They argue that disparities in scale are transient and researchers can work to reduce them . they argue that data, rather than hardware, is still a bottleneck for many applications . |
| Outcome: | a new system trained on well over a trillion words smashes the state of the art by a margin previously thought impossible. |
Cross-lingual Transfer Can Worsen Bias in Sentiment Analysis (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing evidence of demographic bias in SA systems is limited to a handful of languages, and it is costly to create supervised training data in a new language. |
| Approach: | They use counterfactual evaluation to test whether gender or racial biases are imported when using cross-lingual transfer . r&r is much more prevalent than gender biase . |
| Outcome: | The proposed model is compared with monolingual systems in five languages and shows that it is biased more than monolingual ones. |
Bias Beyond English: Counterfactual Tests for Bias in Sentiment Analysis in Four Languages (2023.findings-acl)
Copied to clipboard
| Challenge: | Sentiment analysis systems are used in hundreds of products and languages . Gender and racial biases are well-studied in English, but understudied elsewhere . |
| Approach: | They build a counterfactual evaluation corpus for gender and racial/migrant bias in four languages. |
| Outcome: | The evaluation corpus reveals which models have less bias and pinpoints changes in model bias behaviour, enabling more targeted mitigation strategies. |
Pre-training on high-resource speech recognition improves low-resource speech-to-text translation (N19-1)
Copied to clipboard
| Challenge: | Pre-training on high-resource automatic speech recognition (ASR) tasks improves ST performance even when source language is low-resourced. |
| Approach: | They propose a method to improve direct speech-to-text translation when source language is low-resource . they pre-train model on high-res automatic speech recognition task and fine-tune parameters for ST . |
| Outcome: | The proposed approach improves Spanish English ST even when the source language is low-resource . the pre-trained encoder accounts for most of the improvement, the authors show . |
LSTMs Compose—and Learn—Bottom-Up (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work in NLP shows that LSTMs capture compositional structure in language data. |
| Approach: | They propose to measure the decompositional interdependence between word meanings in an LSTM based on their gate interactions. |
| Outcome: | The proposed model can model syntactic relationships rather than learning the longer-range relations independently. |
Graph Formalisms for Meaning Representations (D18-3)
Copied to clipboard
| Challenge: | In this tutorial we will discuss hyperedge replacement languages (HRL) HRL are a graph rewriting system that is context-free . |
| Approach: | They will discuss hyperedge replacement languages (HRL) they will also discuss Regular Graph Languages (RGL) |
| Outcome: | This tutorial will discuss hyperedge replacement languages (HRL) they are one of the most popular graph formalisms to be studied in NLP . |
Human Temporal Inferences Go Beyond Aspectual Class (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing work on aspectual classification in English has been motivated as a pre-requisite for Natural Language Understanding (NLU) in cases where temporal reasoning is required. |
| Approach: | They propose to classify English verb phrases into situation aspect categories by gathering crowd-sourced judgements from non-expert, native English participants. |
| Outcome: | The proposed approach uses a crowd-sourced dataset from non-expert, native English participants to examine aspectual entailments in English. |
What do character-level models learn about morphology? The case of dependency parsing (D18-1)
Copied to clipboard
| Challenge: | morphologically rich languages require character-level input models to learn morphology, but some models are poor at disambiguating some words . authors of this study show that character- level models learn a lot from input input . explicit modeling of morphologies is expensive and expensive, authors say . |
| Approach: | They compare character-level models to an oracle with explicit morphological analysis . they show that explicitly modeling morphology improves their best model . |
| Outcome: | The results show that character-level models learn morphology better than word models . the authors compare character-based models to oracles on 12 languages with morphological typologies . |
A Structured Syntax-Semantics Interface for English-AMR Alignment (N18-1)
Copied to clipboard
| Challenge: | Abstract Meaning Representation (AMR) annotations do not require explicit mapping between elements of an AMR and the corresponding elements of the sentence that evoke them. |
| Approach: | They devised an expressive framework to align AMR graphs to dependency graphs . their framework explains how 97% of AMR edges are evoked by words or syntax . |
| Outcome: | The proposed framework explains how 97% of AMR edges are evoked by words or syntax. |
World Knowledge Resolves Some Aspectual Ambiguity (2025.findings-acl)
Copied to clipboard
| Challenge: | Annotating event descriptions with aspectual features is often seen as a pre-requisite to temporal reasoning, however, a recent study has shown that non-experts’ annotations of the aspectual class of English verb phrases can disagree with both expert linguistic annotations and each other. |
| Approach: | They hypothesized that people use their world knowledge to tacitly conjure their own contexts, leading to disagreement between them. |
| Outcome: | The results show that the hypothesis explains some of the disagreement, but outputs from GPT-4 are not an accurate predictor of human answers. |
The problem with probabilistic DAG automata for semantic graphs (N19-1)
Copied to clipboard
| Challenge: | Abstract Meaning Representation (AMR) annotations are directed acyclic graphs, but most probabilistic models view them as strings or trees. |
| Approach: | They show that some DAG automata cannot be made into useful probabilistic models by assigning weights to transitions. |
| Outcome: | The proposed model can't be made into useful probabilistic models by assigning weights to transitions . the proposed model is not feasible for all variants, but it is problematic for planar variants if they are not rooted . |
Semantic graph parsing with recurrent neural network DAG grammars (D19-1)
Copied to clipboard
| Challenge: | Semantic parsing is the task of mapping natural language to machine interpretable meaning representations. |
| Approach: | They propose a graph-aware sequence model that generates only well-formed graphs . their model is based on a multilingual semantic graphbank . |
| Outcome: | The proposed model yields competitive results in English and establishes the first results for German, Italian and Dutch. |