Papers by Alexander Fraser
Multilingual Text-to-Image Generation Magnifies Gender Stereotypes (2025.acl-long)
Copied to clipboard
Felix Friedrich, Katharina Hämmerl, Patrick Schramowski, Manuel Brack, Jindřich Libovický, Alexander Fraser, Kristian Kersting
| Challenge: | Text-to-image (T2I) generation models have great results in image quality, flexibility, and text alignment, but they suffer from substantial gender bias. |
| Approach: | They propose a benchmark to study gender bias in multilingual T2I models . they use multilingual prompts to account for grammatical differences influencing gender . |
| Outcome: | The proposed benchmark shows strong gender biases and language-specific differences across models. |
Beyond Literal Token Overlap: Token Alignability for Multilinguality (2025.naacl-short)
Copied to clipboard
| Challenge: | Existing studies have shown that token overlap is a strong predictor of multilinguality and cross-lingual knowledge transfer between languages with different scripts. |
| Approach: | They propose a subword token alignability metric to understand the impact and quality of multilingual tokenisation. |
| Outcome: | The proposed metric predicts multilinguality much better when scripts are disparate and the overlap of literal tokens is low. |
m^4 Adapter: Multilingual Multi-Domain Adaptation for Machine Translation with a Meta-Adapter (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Multilingual neural machine translation models (MNMT) are effective on transferring knowledge between high-resource languages to low-resourced languages. |
| Approach: | They propose a multilingual multi-domain adapter which combines domain and language knowledge using meta-learning with adapters. |
| Outcome: | The proposed model outperforms other adapter methods in a domain shift and language pair translation task. |
Mitigating Data Imbalance and Representation Degeneration in Multilingual Machine Translation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to multilingual neural machine translation (MNMT) are limited in their ability to handle large amounts of data. |
| Approach: | They propose a framework which only requires target-side monolingual data and a bilingual dictionary to improve the performance of the MNMT model. |
| Outcome: | The proposed framework is more effective than baselines in long-tail and high-resource languages. |
Combining Static and Contextualised Multilingual Embeddings (2022.findings-acl)
Copied to clipboard
| Challenge: | Static embeddings are less expressive than contextual language models, but can be more straightforwardly aligned across multiple languages. |
| Approach: | They extract static embeddings for 40 languages from XLM-R and validate them with cross-lingual word retrieval and then align them using VecMap. |
| Outcome: | The proposed approach improves multilingual representations by leveraging static embeddings and a pre-training code. |
Speaking Multiple Languages Affects the Moral Bias of Language Models (2023.findings-acl)
Copied to clipboard
Katharina Haemmerl, Bjoern Deiseroth, Patrick Schramowski, Jindřich Libovický, Constantin Rothkopf, Alexander Fraser, Kristian Kersting
| Challenge: | Pre-trained multilingual language models are often better on English than other languages . however, they are trained on varying amounts of data for each language . |
| Approach: | They apply the MORALDIRECTION framework to multilingual models and analyse their results . they find that PMLMs encode differing moral biases, but these do not correspond to cultural differences or commonalities in human opinions. |
| Outcome: | The proposed model captures moral norms from English and imposes them on other languages. |
Hate Personified: Investigating the role of LLMs in content moderation (2024.emnlp-main)
Copied to clipboard
| Challenge: | Our work provides preliminary guidelines and highlights the nuances of applying Large Language models in culturally sensitive cases. |
| Approach: | They propose to use large language models to help with content moderation to assess how well the needs of diverse groups are reflected in annotated posts. |
| Outcome: | The proposed model is able to leverage community-based flagging efforts and exposure to adversaries. |
ContraCAT: Contrastive Coreference Analytical Templates for Machine Translation (2020.coling-main)
Copied to clipboard
| Challenge: | Recent high scores on pronoun translation suggest current approaches work well . et al., 2018: are context-aware nmt models learning this task? |
| Approach: | They propose a test set to assess the ability to handle specific steps for pronoun translation . they propose heuristics that break down when translations require real reasoning . |
| Outcome: | The proposed model can model complex inferences required for translation of english into german . it shows that current approaches are not able to model all of this information well . |
Improving Low-Resource Languages in Pre-Trained Multilingual Language Models (2022.emnlp-main)
Copied to clipboard
| Challenge: | Pre-trained multilingual language models are the foundation of many NLP approaches, but are often not well-supported by these models due to small available monolingual corpora. |
| Approach: | They propose an unsupervised approach to improve cross-lingual representations of low-resource languages by bootstrapping word translation pairs from monolingual corpora and using them to improve language alignment. |
| Outcome: | The proposed approach improves cross-lingual representations on low-resource languages using word retrieval and zero-shot named entity recognition. |
Data-Efficient Hate Speech Detection via Cross-Lingual Nearest Neighbor Retrieval with Limited Labeled Data (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for detecting hate speech data are expensive and time-consuming . labeled data is expensive and difficult to collect, especially for low-resource languages . |
| Approach: | They propose a method that leverages nearest-neighbor retrieval to augment minimal labeled data in target language. |
| Outcome: | The proposed method outperforms existing models on eight languages and is highly data-efficient. |
Analyzing the Understanding of Morphologically Complex Words in Large Language Models (2024.lrec-main)
Copied to clipboard
| Challenge: | Morphologically complex languages are challenging for NLP as a large amount of information is condensed into a single word, unlike in analytical languages where separate words make it easier to derive meaning. |
| Approach: | They use a Large Language Model to analyse compositional word formation and derivation to find ill-formed word forms. |
| Outcome: | The proposed model is capable of solving most tasks except identifying ill-formed word forms. |
Style-Specific Neurons for Steering LLMs in Text Style Transfer (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing LLMs tend to prioritize preserving original meaning over enhancing stylistic differences in TST. |
| Approach: | They propose a novel approach to steering LLMs using style-specific neurons in TST. |
| Outcome: | Empirical results show that the proposed method improves the fluency of the generated text. |
VOLIMET: A Parallel Corpus of Literal and Metaphorical Verb-Object Pairs for English–German and English–French (2024.starsem-1)
Copied to clipboard
| Challenge: | Metaphorical language is a complex interplay of cultural and linguistic elements that characterizes metaphorical language . a corpus of parallel sentences containing gold standard alignments of metaphorical verb-object pairs and literal paraphrases is presented . |
| Approach: | They propose to analyze metaphorical verb-object pairs and literal paraphrases in parallel sentences from English to German and French. |
| Outcome: | The proposed corpus of 2,916 parallel sentences reveals monolingual patterns for metaphorical vs. literal uses in English . cross-lingually, the results show a rich variability in translations as well as different behaviors for the two target languages . |
Towards Reasonably-Sized Character-Level Transformer NMT by Finetuning Subword Systems (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to train character-level models require very deep architectures that are difficult and slow to train. |
| Approach: | They propose to fine tune a Transformer token-based model to get a model without token segmentation. |
| Outcome: | The proposed model improves translation quality and robustness to noise while requiring less token segmentation. |
Better OOV Translation with Bilingual Terminology Mining (P19-1)
Copied to clipboard
| Challenge: | Unseen words, also called out-of-vocabulary words, are difficult for machine translation . byte-pair encoding can be used to represent OOVs, but they are often incorrectly translated . |
| Approach: | They propose to use monolingual data to improve the translation of unseen words . they use five target language words to mine target-language sentences . |
| Outcome: | The proposed system can be used to improve translation of out-of-vocabulary words (OOVs) the proposed system is trained on Europarl and can be fine-tuned to improve the translation quality. |
CUTE: Measuring LLMs’ Understanding of Their Tokens (2024.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) perform well on a wide variety of tasks, authors say . they lack direct access to characters, which can be difficult to generalize to new languages . |
| Approach: | They propose a benchmark to test the orthographic knowledge of Large Language Models . they find that most LLMs seem to know the spelling of their tokens - yet fail to manipulate text . |
| Outcome: | The proposed benchmark tests the orthographic knowledge of large language models . it finds that most LLMs seem to know the spelling of their tokens, but fail to manipulate text . |
EmoBench-UA: A Benchmark Dataset for Emotion Detection in Ukrainian (2025.findings-emnlp)
Copied to clipboard
| Challenge: | **EmoBench-UA** is the first annotated dataset for emotion classification in Ukrainian texts. |
| Approach: | They introduce **EmoBench-UA**, the first annotated dataset for emotion detection in Ukrainian texts. |
| Outcome: | The first annotated dataset for emotion detection in Ukrainian texts is presented in this paper . the dataset was created through crowdsourcing using the Toloka.ai platform . |
LLMs Beyond English: Scaling the Multilingual Capability of LLMs with Cross-Lingual Feedback (2024.findings-acl)
Copied to clipboard
| Challenge: | Recent multilingual models support limited number of human languages due to lack of training data for low resource languages. |
| Approach: | They propose a multilingual multilingual LLM that scales to 100 languages . they use a human feedback dataset and a data set to perform multilingual instruction tuning . |
| Outcome: | The proposed model outperforms its peers on five multilingual benchmarks. |
From Unaligned to Aligned: Scaling Multilingual LLMs with Multi-Way Parallel Corpora (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experiments show that models trained on multi-way parallel data outperform those trained on unaligned data. |
| Approach: | They propose a large-scale, high-quality multi-way parallel corpus based on TED Talks that spans 113 languages with up to 50 languages aligned in parallel. |
| Outcome: | The proposed model outperforms models trained on unaligned multilingual data on six multilingual benchmarks. |
Why don’t people use character-level machine translation? (2022.findings-acl)
Copied to clipboard
| Challenge: | despite evidence character-level systems are comparable with subword systems, they are rarely used in competitive setups in machine translation competitions. |
| Approach: | They propose a two-step decoder architecture that does not suffer from a slow-down due to the length of character sequences. |
| Outcome: | The proposed character-level MT systems show better domain robustness and better morphological generalization . the proposed decoder architecture shows no slow-down due to the length of character sequences . |
Improving Parallel Sentence Mining for Low-Resource and Endangered Languages (2025.acl-short)
Copied to clipboard
| Challenge: | Parallel sentence mining is a technique used to find matching sentence pairs from a source and target language. |
| Approach: | They propose a benchmark dataset for parallel sentence mining on three low-resource languages . they apply alignment post-processing and cluster-based isotropy enhancement techniques to one of them . |
| Outcome: | The proposed datasets show better mining quality overall for low-resource languages . the proposed methods are crucial for optimizing parallel data extraction for low resource languages - a new study shows. |
Improving Both Domain Robustness and Domain Adaptability in Machine Translation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing approaches to domain adaptation for NMT depend on high-quality parallel data. |
| Approach: | They propose a meta-learning framework which improves domain robustness and adaptability . they use a word-level domain mixing model and a domain classifier to integrate it . |
| Outcome: | The proposed approach improves domain robustness and adaptability in seen and unseen domains. |
A Study on Accessing Linguistic Information in Pre-Trained Language Models by Using Prompts (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to access linguistic information in pre-trained multilingual language models are difficult to use. |
| Approach: | They propose prompting and formulate linguistic tasks to test the LM's access to explicit grammar principles and find out what type of information can be obtained . |
| Outcome: | The proposed method can provide access to linguistic features in pre-trained models, but some are harder to capture . |
Adapting Entities across Languages and Cultures (2021.findings-emnlp)
Copied to clipboard
| Challenge: | a structured knowledge base adapts named entities using their shared properties. |
| Approach: | They propose automatic methods to adapt named entities using shared properties . they compare them to human adaptations using a new dataset of human adaptation data . |
| Outcome: | The proposed methods compare to human adaptations using a new dataset. |
Extracting Linguistic Information from Large Language Models: Syntactic Relations and Derivational Knowledge (2025.emnlp-main)
Copied to clipboard
| Challenge: | Using large language models, we study their morphosyntactic competence and generalization capabilities. |
| Approach: | They propose to use morphosyntactic tasks to study their linguistic knowledge and generalization capabilities to extract different types of morphological structure for typologically diverse languages. |
| Outcome: | The proposed models outperform GPT-4o and LLaMA 3.3-70B in all diagnostic tasks, but show little evidence of abstract morphological rule learning. |
Improving the Lexical Ability of Pretrained Language Models for Unsupervised Neural Machine Translation (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing methods for unsupervised neural machine translation (UNMT) use cross-lingual pretraining to align the lexical- and high-level representations of two languages. |
| Approach: | They propose to use type-level cross-lingual subword embeddings to enhance the bilingual masked language model pretraining with lexical-level information to align the two languages. |
| Outcome: | Empirical results show that the method improves on UNMT (up to 4.5 BLEU) and bilingual lexicon induction compared to baseline models. |
EXECUTE: A Multilingual Benchmark for LLM Token Understanding (2025.findings-acl)
Copied to clipboard
| Challenge: | EXECUTE is an expandable X(Cross)-Lingual Extension of CUTE that can be expanded to any language. |
| Approach: | They extend the CUTE benchmark to more languages with diverse scripts and writing systems, introducing EXECUTE. |
| Outcome: | The extended framework allows expansion to any language. |
Positional Overload: Positional Debiasing and Context Window Extension for Large Language Models using Set Encoding (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models typically track the order of tokens using positional encoding, which causes two significant limitations: 1. Positional Bias: When processing long text sequences, the number of token can exceed the range the model was trained on. |
| Approach: | They propose a method that allows multiple pieces of text to be encoded in the same position, eliminating positional bias entirely. |
| Outcome: | The proposed method eliminates positional bias entirely and increases the size of the input an LLM can handle. |
Exploring Bilingual Word Embeddings for Hiligaynon, a Low-Resource Language (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing studies on Hiligaynon, a low-resource language of Malayo-Polynesian origin, have not explored the use of bilingual word embeddings in NLP. |
| Approach: | They use a publicly available Hiligaynon corpus with only 300K words to match it with a comparable English corpus. |
| Outcome: | The proposed model outperforms results from a low-resource language of Malayo-Polynesian origin with over 9 million speakers in the Philippines. |
How to Solve Few-Shot Abusive Content Detection Using the Data We Actually Have (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing datasets for abusive language detection are expensive and lack of knowledge about the target is a challenge. |
| Approach: | They propose to build models cheaply for a new target label set and/or language, using only a few training examples of the target domain. |
| Outcome: | The proposed model improves monolingually and across languages using existing datasets and only a few-shots of the target domain. |
Combining Word Embeddings with Bilingual Orthography Embeddings for Bilingual Dictionary Induction (2020.coling-main)
Copied to clipboard
| Challenge: | Bilingual dictionary induction (BDI) is a task of finding target language translations of source language words. |
| Approach: | They propose to use bilingual orthography Embeddings to enrich BWE-based BDI with transliteration information to make a decision on which information source is more reliable for a particular word pair. |
| Outcome: | The proposed system improves on English-Russian BDI and shows that it can be built with only weak bilingual signals and even without any bilingual signal. |
AdapterSoup: Weight Averaging to Improve Generalization of Pretrained Language Models (2023.findings-eacl)
Copied to clipboard
| Challenge: | Pretrained language models often need to specialize to specific domains. |
| Approach: | They propose an approach that performs weight-space averaging of adapters trained on different domains. |
| Outcome: | The proposed approach improves performance to new domains without extra training. |
Subword Segmentation in LLMs: Looking at Inflection and Consistency (2024.emnlp-main)
Copied to clipboard
| Challenge: | Subword segmentation is not linguistically guided and is not currently well understood in LLMs. |
| Approach: | They group words according to their segmentation properties and compare how well a model can solve a linguistic task for these groups using two criteria: adherence to morpheme boundaries and segmentation consistency of inflected forms of a lemma. |
| Outcome: | The results show that the criterion of segmentation consistency can predict the model’s ability to recognize and generate the lemma from an inflected form, providing evidence that subword segmentation is relevant. |
Evaluating bilingual word embeddings on the long tail (N18-2)
Copied to clipboard
| Challenge: | Bilingual word embeddings are useful for bilingual lexicon induction, but they focus on frequent words in general domains. |
| Approach: | They propose to evaluate bilingual word embeddings on rare words in different domains . they propose to use a multilingual dataset to build and combine BWEs based on a single word . |
| Outcome: | The proposed evaluations show that state-of-the-art methods fail on rare words . the proposed evaluation is based on a gold standard dataset and code . |
Modeling Word Formation in English–German Neural Machine Translation (2020.acl-main)
Copied to clipboard
| Challenge: | linguistically sound word segmentation approaches overcome word formation problems . word-level approaches to MT lack morphological generalization for large vocabulary . |
| Approach: | They propose a word segmentation approach that considers fusional morphology to model word formation . they apply a linguistically sound segmentation method to both the source and target sides . |
| Outcome: | The proposed approach overcomes the problems caused by fusional morphology . the best system variants employ source-side morphological analysis and model complex target-side words . |
LLM Sensitivity Challenges in Abusive Language Detection: Instruction-Tuned vs. Human Feedback (2025.coling-main)
Copied to clipboard
| Challenge: | Existing studies show that instruction-tuned LLMs under-predict positive classes . however, they are overly sensitive and can be applied for abuse detection without fine-tuning . |
| Approach: | They show that instruction-tuned LLMs tend to under-predict positive classes . they also show that label frequency in the prompt helps with the significant over-prediction . |
| Outcome: | The proposed models under-predict positive classes in social media, whereas they are overly sensitive. |
Unsupervised Parallel Sentence Extraction with Parallel Segment Detection Helps Machine Translation (P19-1)
Copied to clipboard
| Challenge: | Recent advances in unsupervised bilingual word embeddings make it possible to mine parallel sentences from comparable corpora. |
| Approach: | They propose a strong unsupervised system for parallel sentence mining based on cosine similarities of source and target words . they show that parallel sentences mined from real-life sources improve unsupervised MT . |
| Outcome: | The proposed system improves unsupervised MT on three language pairs. |
Two Methods for Domain Adaptation of Bilingual Tasks: Delightfully Simple and Broadly Applicable (P18-1)
Copied to clipboard
| Challenge: | Previously, domain adaptation approaches to bilingual tasks were proposed . we show that simple adaptation process involving only unlabeled text is highly effective . |
| Approach: | They propose a method for domain adaptation of bilingual word embeddings using unlabeled data . they then tailor a semi-supervised classification method from computer vision to these tasks . |
| Outcome: | The proposed method improves on two bilingual tasks using unlabeled data. |
Optical Character Recognition for the International Phonetic Alphabet (2026.eacl-short)
Copied to clipboard
| Challenge: | Grammar books are increasingly used as additional reference resources for low-resource languages . a significant portion of these documents come from scans and require an OCR tool . |
| Approach: | They compare two neural OCR frameworks and a large vision-language model with a synthetic dataset based on Wiktionary to study the International Phonetic Alphabet (IPA). |
| Outcome: | The proposed model improves on the International Phonetic Alphabet (IPA) character set. |
On the Language Neutrality of Pre-trained Multilingual Representations (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies have focused on cross-linguality of contextual embeddings . however, they are only moderately language-neutral by default . |
| Approach: | They propose to use unsupervised centering and fitting an explicit projection on parallel data to achieve stronger language neutrality. |
| Outcome: | The proposed model outperforms existing models on XNLI and NER tasks. |
Embedding Learning Through Multilingual Concept Induction (P18-1)
Copied to clipboard
| Challenge: | Existing methods for learning vector space representations of words are based on word-context information. |
| Approach: | They propose a method for estimating vector space representations of words by concept induction. |
| Outcome: | The proposed method performs better on crosslingual word similarity and sentiment analysis on a parallel corpus. |
NLP for Social Good: A Survey and Outlook of Challenges, Opportunities and Responsible Deployment (2026.eacl-long)
Copied to clipboard
Antonia Karamolegkou, Angana Borah, Eunjung Cho, Sagnik Ray Choudhury, Martina Galletti, Pranav Gupta, Oana Ignat, Priyanka Kargupta, Neema Kotonya, Hemank Lamba, Sun-Joo Lee, Arushi Mangla, Ishani Mondal, Fatima Zahra Moudakir, Deniz Nazar, Poli Nemkova, Dina Pisarevskaya, Naquee Rizwan, Nazanin Sabri, Keenan Samway, Dominik Stammbach, Anna Steinberg Schulten, David Tomás, Steven R Wilson, Bowen Yi, Jessica H Zhu, Arkaitz Zubiaga, Anders Søgaard, Alexander Fraser, Zhijing Jin, Rada Mihalcea, Joel R. Tetreault, Daryna Dementieva
| Challenge: | This paper surveys work in "NLP for Social Good" across nine domains relevant to global development and risk agendas. |
| Approach: | This paper analyzes work in "NLP for Social Good" across nine domains relevant to global development and risk agendas. |
| Outcome: | The paper analyzes work in "NLP for Social Good" across nine domains relevant to global development and risk agendas. |
Reusing a Pretrained Language Model on Languages with Limited Corpora for Unsupervised NMT (2020.emnlp-main)
Copied to clipboard
| Challenge: | Neural machine translation (NMT) models with limited data are ineffective when the two languages are not available for one language. |
| Approach: | They propose an approach that reuses a language model that is pretrained on two languages with large monolingual data to initialize an unsupervised neural machine translation system. |
| Outcome: | The proposed method outperforms a competitive cross-lingual pretraining model in English-Macedonian (En-Mk) and English-Albanian (En Sq) it yields more than +8.3 BLEU points for all four translation directions. |
A Survey of Methods for Addressing Class Imbalance in Deep-Learning Based Natural Language Processing (2023.eacl-main)
Copied to clipboard
| Challenge: | Developing methods to improve model performance in imbalanced data settings has been an active area for decades . |
| Approach: | They propose to use sampling, data augmentation, choice of loss function, staged learning, or model design to address class imbalance in NLP. |
| Outcome: | The proposed approaches are evaluated on a variety of NLP tasks or in the computer vision community. |
Reconstruction of Cuneiform Literary Texts as Text Matching (2024.lrec-main)
Copied to clipboard
| Challenge: | cuneiform fragment identification is a slow and unsystematic process for reconstructing ancient texts . fragments of cuniform script are often found in fragments written on clay tablets . cnl is able to identify fragments and match them with existing text collections . |
| Approach: | They propose a character-level n-gram-based similarity matching approach to identify fragments . they compare different approaches to identify overlaps between fragments and texts . |
| Outcome: | The proposed approach speeds up the process and reduces the time it takes to complete the work. |
Exploring Anisotropy and Outliers in Multilingual Language Models for Cross-Lingual Semantic Sentence Similarity (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have shown that contextual language models display outlier dimensions . this is true for monolingual and multilingual models, but little work has been done on multilingual contexts . |
| Approach: | They investigate outlier dimensions and their relationship to anisotropy in multilingual contexts . they focus on cross-lingual semantic similarity tasks . |
| Outcome: | The proposed model improves on cross-lingual semantic similarity tasks. |
Fine-Grained Transfer Learning for Harmful Content Detection through Label-Specific Soft Prompt Tuning (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing detection models are less effective and generalizable due to static data. |
| Approach: | They propose a method that leverages class-specific knowledge to enhance harmful content detection. |
| Outcome: | The proposed method improves harmful content detection across English and German datasets. |
Anchor-based Bilingual Word Embeddings for Low-Resource Languages (2021.acl-short)
Copied to clipboard
| Challenge: | Existing approaches to build monolingual word embeddings rely on a cheap bilingual signal and monolingual data. |
| Approach: | They propose a method where the vector space of the high resource source language is used as a starting point for training an embedding space for the low resource target language. |
| Outcome: | The proposed approach improves bilingual lexicon induction performance and target language MWE quality. |
Understanding Cross-Lingual Alignment—A Survey (2024.findings-acl)
Copied to clipboard
| Challenge: | Cross-lingual alignment is the meaningful similarity of representations across languages in multilingual language models. |
| Approach: | They propose a taxonomy of methods to improve cross-lingual alignment . they argue that an effective trade-off between language-neutral and language-specific information is key . |
| Outcome: | The proposed methods can be applied to encoder models and encoder-decoder-only models . they show that language-neutral and language-specific information is key . |
Dial HEALTHDIAL for Advice: A Multilingual and Multi-Parallel Spoken Dialogue Dataset for Knowledge-Grounded Information Seeking (2026.findings-acl)
Copied to clipboard
Songbo Hu, Yinhong Liu, Ej Zhou, Evgeniia Razumovskaia, Xiaobin Wang, Alexander Fraser, Ivan Vulić, Anna Korhonen
| Challenge: | Creating spoken dialogue datasets is methodologically challenging due to the personally identifiable nature of speech signals. |
| Approach: | They propose a large-scale, multilingual, and multi-parallel dataset for developing and evaluating retrieval-augmented generation-based spoken dialogue systems. |
| Outcome: | The proposed dataset includes 6,000 information-seeking dialogues and 163 hours of user speech recorded from native speakers of four official WHO languages. |