Papers by Ákos Kádár
TURING: an Accurate and Interpretable Multi-Hypothesis Cross-Domain Natural Language Database Interface (2021.acl-demo)
Copied to clipboard
Peng Xu, Wenjie Zi, Hamidreza Shahidi, Ákos Kádár, Keyi Tang, Wei Yang, Jawad Ateeq, Harsh Barot, Meidan Alon, Yanshuai Cao
| Challenge: | Existing text-to-SQL semantic parsers cannot achieve high accuracy in cross-database setting . TURING is a NLDB system that can be used to democratize data-driven insights for non-technical users . |
| Approach: | They propose a TURING system that provides high-precision natural language explanations of SQL queries in a beam. |
| Outcome: | The proposed system achieves 75.1% execution accuracy and 78.3% top-5 beam execution accuracy on the Spider validation set. |
DIDEC: The Dutch Image Description and Eye-tracking Corpus (C18-1)
Copied to clipboard
| Challenge: | Using a corpus of spoken Dutch image descriptions and eye-tracking data, we can gain a deeper understanding of the image description task, especially how visual attention is correlated with the image descriptions. |
| Approach: | They present a corpus of spoken Dutch image descriptions paired with eye-tracking data from Free viewing and Description viewing tasks to provide an initial analysis of self-corrections in image descriptions. |
| Outcome: | The results show that the eye-tracking data for the description viewing task is more coherent than for the free-viewing task, and that variation in image descriptions is only moderately correlated across different languages. |
Dependency parsing with structure preserving embeddings (2021.eacl-main)
Copied to clipboard
| Challenge: | Modern neural approaches to dependency parsing are trained to predict a tree structure by learning a contextual representation for tokens in a sentence and a head–dependent scoring function. |
| Approach: | They propose to combine a contextual representation for tokens and a head–dependent scoring function to learn interpretable representations by training a parser to explicitly preserve structural properties of a tree. |
| Outcome: | The proposed approach yields strong tree distance preservation and parsing performance on par with a competitive graph-based parser. |
Revisiting the Hierarchical Multiscale LSTM (C18-1)
Copied to clipboard
| Challenge: | Hierarchical Multiscale LSTM model learns structure from character input . high complexity of architecture, training and implementations might hinder its applicability . |
| Approach: | They propose to reproduce and ablate hierarchical multiscale LSTM language model and show that simplifying certain aspects of the architecture can improve its performance. |
| Outcome: | The proposed model performs better when simplified and linguistic units are learned by different levels of the model. |
Adversarial Stylometry in the Wild: Transferable Lexical Substitution Attacks on Author Profiling (2021.eacl-main)
Copied to clipboard
| Challenge: | Written language contains stylistic cues that can be exploited to automatically infer a variety of potentially sensitive author information. |
| Approach: | They propose to use a transformer-based extension of a lexical replacement attack to attack written language by rewriting an author's text. |
| Outcome: | The proposed framework achieves high transferability when trained on a weakly labeled corpus—decreasing target model performance below chance. |
Accurate polyglot semantic parsing with DAG grammars (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Semantic parsers treat graphs as strings or trees, but there is no guarantee that the output is a well-formed graph. |
| Approach: | They propose a graph-aware sequence model that utilizes a DAG grammar to guide graph generation. |
| Outcome: | The proposed model outperforms string-based and DAG-grammar models by a large margin and can guarantee the well-formed graphs. |
Subword Pooling Makes a Difference (2021.eacl-main)
Copied to clipboard
| Challenge: | Contextual word-representations use subword tokenization to handle large vocabularies and unknown words. |
| Approach: | They propose to use the first subword for morphological probing, POS tagging and NER to pool multiple subwords that correspond to a single word in contextual language models. |
| Outcome: | The proposed model outperforms two multilingual models on morphological probing, POS tagging and NER tasks in 9 languages. |
Cyberbullying Classifiers are Sensitive to Model-Agnostic Perturbations (2022.lrec-1)
Copied to clipboard
| Challenge: | toxicity classifiers rely on lexical cues, so creative language use can be detrimental to utility of current corpora and state-of-the-art models. |
| Approach: | They propose to use model-agnostic adversarial behavior to enhance toxic content classification models. |
| Outcome: | The proposed model-agnostic adversarial behavior and augmentation for cyberbullying detection are robust against word-level perturbations at a slight trade-off in overall task performance. |
Improving Lemmatization of Non-Standard Languages with Joint Learning (N19-1)
Copied to clipboard
| Challenge: | Lemmatization is a task of mapping a token to its corresponding dictionary head-form to abstract away from orthographic and inflectional variation. |
| Approach: | They propose to improve lemmatization performance on non-standard historical languages . they propose an Encoder-Decoder architecture which enriches with sentence information . |
| Outcome: | The proposed model does not require POS or morphological annotations, which are not always available for historical corpora. |
NeuralREG: An end-to-end approach to referring expression generation (P18-1)
Copied to clipboard
| Challenge: | Referring Expression Generation models typically rely on features such as salience and grammatical function to make decisions about form and content. |
| Approach: | They propose a new approach that makes decisions about form and content in one go . they use a delexicalized version of the WebNLG corpus to test the approach . |
| Outcome: | The proposed approach significantly improves over two strong baselines. |