Cross-lingual transfer has become a crucial aspect of multilingual NLP, as it allows for models trained on resource-rich languages to be applied to low-resource languages more effectively. Recently massively multilingual pre-trained language models (e.g., mBERT, XLM-R) demonstrate strong zero-shot transfer capabilities[14] [13]. This paper investigates cross-linguistic transfer through the lens of language families and morphology. Investigating how language family proximity and morphological similarity affect performance across NLP tasks. We further discuss our results and how it relates to findings from recent literature. Overall, we compare multilingual model performance and review how linguistic distance metrics correlate with transfer outcomes. We also look into emerging approaches that integrate typological and morphological information into model pre-training to improve transfer to diverse languages[18] [19].
Accurate detection of offensive language is essential for a number of applications related to social media safety. There is a sharp contrast in performance in this task between low and high-resource languages. In this paper, we adapt fine-tuning strategies that have not been previously explored for Sinhala in the downstream task of offensive language detection. Using this approach, we introduce four models: "Subasa-XLM-R", which incorporates an intermediate Pre-Finetuning step using Masked Rationale Prediction. Two variants of "Subasa-Llama" and "Subasa-Mistral", are fine-tuned versions of Llama (3.2) and Mistral (v0.3), respectively, with a task-specific strategy. We evaluate our models on the SOLD benchmark dataset for Sinhala offensive language detection. All our models outperform existing baselines. Subasa-XLM-R achieves the highest Macro F1 score (0.84) surpassing state-of-the-art large language models like GPT-4o when evaluated on the same SOLD benchmark dataset under zero-shot settings. The models and code are publicly available.
While multilingual language models like XLM-R have advanced multilingualism in NLP, they still perform poorly in extremely low-resource languages. This situation is exacerbated by the fact that modern LLMs such as LLaMA and Qwen support far fewer languages than XLM-R, making text generation models non-existent for many languages in the world. To tackle this challenge, we propose a novel framework for adapting multilingual encoders to text generation in extremely low-resource languages. By reusing the weights between the encoder and the decoder, our framework allows the model to leverage the learned semantic space of the encoder, enabling efficient learning and effective generalization in low-resource languages. Applying this framework to four Chinese minority languages, we present XLM-SWCM, and demonstrate its superior performance on various downstream tasks even when compared with much larger models.
This research presents a novel framework for translating extractive question-answering datasets into low-resource languages, as demonstrated by the creation of the AmaSQuAD dataset, a translation of SQuAD 2.0 into Amharic. The methodology addresses challenges related to misalignment between translated questions and answers, as well as the presence of multiple answer instances in the translated context. For this purpose, we used cosine similarity utilizing embeddings from a fine-tuned BERT-based model for Amharic and Longest Common Subsequence (LCS). Additionally, we fine-tune the XLM-R model on the AmaSQuAD synthetic dataset for Amharic Question-Answering. The results show an improvement in baseline performance, with the fine-tuned model achieving an increase in the F1 score from 36.55% to 44.41% and 50.01% to 57.5% on the AmaSQuAD development dataset. Moreover, the model demonstrates improvement on the human-curated AmQA dataset, increasing the F1 score from 67.80% to 68.80% and the exact match score from 52.50% to 52.66%.The AmaSQuAD dataset is publicly available Datasets
The black-box nature of large language models (LLMs) necessitates the development of eXplainable AI (XAI) techniques for transparency and trustworthiness. However, evaluating these techniques remains a challenge. This study presents a general evaluation framework using four key metrics: Human-reasoning Agreement (HA), Robustness, Consistency, and Contrastivity. We assess the effectiveness of six explainability techniques from five different XAI categories model simplification (LIME), perturbation-based methods (SHAP), gradient-based approaches (InputXGradient, Grad-CAM), Layer-wise Relevance Propagation (LRP), and attention mechanisms-based explainability methods (Attention Mechanism Visualization, AMV) across five encoder-based language models: TinyBERT, BERTbase, BERTlarge, XLM-R large, and DeBERTa-xlarge, using the IMDB Movie Reviews and Tweet Sentiment Extraction (TSE) datasets. Our findings show that the model simplification-based XAI method (LIME) consistently outperforms across multiple metrics and models, significantly excelling in HA with a score of 0.9685 on DeBERTa-xlarge, robustness, and consistency as the complexity of large language models increases. AMV demonstrates the best Robustness, with scores as low as 0.0020. It also excels in Consistency, achieving near-perfect scores of 0.9999 across all models. Regarding Contrastivity, LRP performs the best, particularly on more complex models, with scores up to 0.9371.
This study explores transformer-based models such as BERT, mBERT, and XLM-R for multi-lingual sentiment analysis across diverse linguistic structures. Key contributions include the identification of XLM-R superior adaptability in morphologically complex languages, achieving accuracy levels above 88%. The work highlights fine-tuning strategies and emphasizes their significance for improving sentiment classification in underrepresented languages.
This paper presents results of our system for CoMeDi Shared Task, focusing on Subtask 2: Disagreement Ranking. Our system leverages sentence embeddings generated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep neural regression model incorporating batch normalization and dropout for improved generalization. By predicting the mean of pairwise judgment differences between annotators, our method explicitly targets disagreement ranking, diverging from traditional "gold label" aggregation approaches. We optimized our system with a customized architecture and training procedure, achieving competitive performance in Spearman correlation against mean disagreement labels. Our results highlight the importance of robust embeddings, effective model architecture, and careful handling of judgment differences for ranking disagreement in multilingual contexts. These findings provide insights into the use of contextualized representations for ordinal judgment tasks and open avenues for further refinement of disagreement prediction models.
This study examines the use of Natural Language Processing (NLP) technology within the Islamic domain, focusing on developing an Islamic neural retrieval model. By leveraging the robust XLM-R model, the research employs a language reduction technique to create a lightweight bilingual large language model (LLM). Our approach for domain adaptation addresses the unique challenges faced in the Islamic domain, where substantial in-domain corpora exist only in Arabic while limited in other languages, including English. The work utilizes a multi-stage training process for retrieval models, incorporating large retrieval datasets, such as MS MARCO, and smaller, in-domain datasets to improve retrieval performance. Additionally, we have curated an in-domain retrieval dataset in English by employing data augmentation techniques and involving a reliable Islamic source. This approach enhances the domain-specific dataset for retrieval, leading to further performance gains. The findings suggest that combining domain adaptation and a multi-stage training method for the bilingual Islamic neural retrieval model enables it to outperform monolingual models on downstream retrieval tasks.
The BabyLM challenge called on participants to develop sample-efficient language models. Submissions were pretrained on a fixed English corpus, limited to the amount of words children are exposed to in development (<100m). The challenge produced new architectures for data-efficient language modelling, which outperformed models trained on trillions of words. This is promising for low-resource languages, where available corpora are limited to much less than 100m words. In this paper, we explore the potential of BabyLMs for low-resource languages, using the isiXhosa language as a case study. We pretrain two BabyLM architectures, ELC-BERT and MLSM, on an isiXhosa corpus. They outperform a vanilla pretrained model on POS tagging and NER, achieving notable gains (+3.2 F1) for the latter. In some instances, the BabyLMs even outperform XLM-R. Our findings show that data-efficient models are viable for low-resource languages, but highlight the continued importance, and lack of, high-quality pretraining data. Finally, we visually analyse how BabyLM architectures encode isiXhosa.
Cross-lingual semantic textual relatedness task is an important research task that addresses challenges in cross-lingual communication and text understanding. It helps establish semantic connections between different languages, crucial for downstream tasks like machine translation, multilingual information retrieval, and cross-lingual text understanding.Based on extensive comparative experiments, we choose the XLM-R-base as our base model and use pre-trained sentence representations based on whitening to reduce anisotropy.Additionally, for the given training data, we design a delicate data filtering method to alleviate the curse of multilingualism. With our approach, we achieve a 2nd score in Spanish, a 3rd in Indonesian, and multiple entries in the top ten results in the competition's track C. We further do a comprehensive analysis to inspire future research aimed at improving performance on cross-lingual tasks.
Current language models (LMs) use a fixed, static subword tokenizer. This default choice typically results in degraded efficiency and language capabilities, especially in languages other than English. To address this issue, we challenge the static design and propose retrofitting LMs with dynamic tokenization: a way to dynamically decide on token boundaries based on the input text via a subword-merging algorithm inspired by byte-pair encoding. We merge frequent subword sequences in a batch, then apply a pre-trained embedding-prediction hypernetwork to compute the token embeddings on-the-fly. For encoder-style models (e.g., XLM-R), this on average reduces token sequence lengths by >20% across 14 languages while degrading performance by less than 2%. The same method applied to pre-filling and scoring in decoder-style models (e.g., Mistral-7B; evaluated on English) results in minimal performance degradation at up to 6% reduction in sequence length. Overall, we find that dynamic tokenization can mitigate the limitations of static tokenization by substantially improving inference speed and promoting fairness across languages, enabling more equitable and adaptable LMs.
Natural Language Inference (NLI) is a task within Natural Language Processing (NLP) that holds value for various AI applications. However, there have been limited studies on Natural Language Inference in Vietnamese that explore the concept of joint models. Therefore, we conducted experiments using various combinations of contextualized language models (CLM) and neural networks. We use CLM to create contextualized work presentations and use Neural Networks for classification. Furthermore, we have evaluated the strengths and weaknesses of each joint model and identified the model failure points in the Vietnamese context. The highest F1 score in this experiment, up to 82.78% in the benchmark dataset (ViNLI). By conducting experiments with various models, the most considerable size of the CLM is XLM-R (355M). That combination has consistently demonstrated superior performance compared to fine-tuning strong pre-trained language models like PhoBERT (+6.58%), mBERT (+19.08%), and XLM-R (+0.94%) in terms of F1-score. This article aims to introduce a novel approach or model that attains improved performance for Vietnamese NLI. Overall, we find that the joint approach of CLM and neural networks is simple yet capable of achieving high-quality performance, which makes it suitable for applications that require efficient resource utilization.
In this paper, we investigate the use of N-gram models and Large Pre-trained Multilingual models for Language Identification (LID) across 11 South African languages. For N-gram models, this study shows that effective data size selection remains crucial for establishing effective frequency distributions of the target languages, that efficiently model each language, thus, improving language ranking. For pre-trained multilingual models, we conduct extensive experiments covering a diverse set of massively pre-trained multilingual (PLM) models -- mBERT, RemBERT, XLM-r, and Afri-centric multilingual models -- AfriBERTa, Afro-XLMr, AfroLM, and Serengeti. We further compare these models with available large-scale Language Identification tools: Compact Language Detector v3 (CLD V3), AfroLID, GlotLID, and OpenLID to highlight the importance of focused-based LID. From these, we show that Serengeti is a superior model across models: N-grams to Transformers on average. Moreover, we propose a lightweight BERT-based LID model (za_BERT_lid) trained with NHCLT + Vukzenzele corpus, which performs on par with our best-performing Afri-centric models.
Recent multilingual pretrained language models (mPLMs) often avoid using language embeddings -- learnable vectors assigned to different languages. These embeddings are discarded for two main reasons: (1) mPLMs are expected to have a single, unified parameter set across all languages, and (2) they need to function seamlessly as universal text encoders without requiring language IDs as input. However, this removal increases the burden on token embeddings to encode all language-specific information, which may hinder the model's ability to produce more language-neutral representations. To address this challenge, we propose Language-Script Aware Multilingual Pretraining (LangSAMP), a method that incorporates both language and script embeddings to enhance representation learning while maintaining a simple architecture. Specifically, we integrate these embeddings into the output of the transformer blocks before passing the final representations to the language modeling head for prediction. We apply LangSAMP to the continual pretraining of XLM-R on a highly multilingual corpus covering more than 500 languages. The resulting model consistently outperforms the baseline. Extensive analysis further shows that language/script embeddings encode language/script-specific information, which improves the selection of source languages for crosslingual transfer. We make our code and models publicly available at \url{https://github.com/cisnlp/LangSAMP}.
Contextualized embeddings based on large language models (LLMs) are available for various languages, but their coverage is often limited for lower resourced languages. Using LLMs for such languages is often difficult due to a high computational cost; not only during training, but also during inference. Static word embeddings are much more resource-efficient ("green"), and thus still provide value, particularly for very low-resource languages. There is, however, a notable lack of comprehensive repositories with such embeddings for diverse languages. To address this gap, we present GrEmLIn, a centralized repository of green, static baseline embeddings for 87 mid- and low-resource languages. We compute GrEmLIn embeddings with a novel method that enhances GloVe embeddings by integrating multilingual graph knowledge, which makes our static embeddings competitive with LLM representations, while being parameter-free at inference time. Our experiments demonstrate that GrEmLIn embeddings outperform state-of-the-art contextualized embeddings from E5 on the task of lexical similarity. They remain competitive in extrinsic evaluation tasks like sentiment analysis and natural language inference, with average performance gaps of just 5-10\% or less compared to state-of-the-art models, given a sufficient vocabulary overlap with the target task, and underperform only on topic classification. Our code and embeddings are publicly available at https://huggingface.co/DFKI.
We present systematic efforts in building long-context multilingual text representation model (TRM) and reranker from scratch for text retrieval. We first introduce a text encoder (base size) enhanced with RoPE and unpadding, pre-trained in a native 8192-token context (longer than 512 of previous multilingual encoders). Then we construct a hybrid TRM and a cross-encoder reranker by contrastive learning. Evaluations show that our text encoder outperforms the same-sized previous state-of-the-art XLM-R. Meanwhile, our TRM and reranker match the performance of large-sized state-of-the-art BGE-M3 models and achieve better results on long-context retrieval benchmarks. Further analysis demonstrate that our proposed models exhibit higher efficiency during both training and inference. We believe their efficiency and effectiveness could benefit various researches and industrial applications.
Sentiment analysis serves as a pivotal component in Natural Language Processing (NLP). Advancements in multilingual pre-trained models such as XLM-R and mT5 have contributed to the increasing interest in cross-lingual sentiment analysis. The recent emergence in Large Language Models (LLM) has significantly advanced general NLP tasks, however, the capability of such LLMs in cross-lingual sentiment analysis has not been fully studied. This work undertakes an empirical analysis to compare the cross-lingual transfer capability of public Small Multilingual Language Models (SMLM) like XLM-R, against English-centric LLMs such as Llama-3, in the context of sentiment analysis across English, Spanish, French and Chinese. Our findings reveal that among public models, SMLMs exhibit superior zero-shot cross-lingual performance relative to LLMs. However, in few-shot cross-lingual settings, public LLMs demonstrate an enhanced adaptive potential. In addition, we observe that proprietary GPT-3.5 and GPT-4 lead in zero-shot cross-lingual capability, but are outpaced by public models in few-shot scenarios.
Spoken Named Entity Recognition (NER) aims to extract named entities from speech and categorise them into types like person, location, organization, etc. In this work, we present VietMed-NER - the first spoken NER dataset in the medical domain. To our knowledge, our Vietnamese real-world dataset is the largest spoken NER dataset in the world regarding the number of entity types, featuring 18 distinct types. Furthermore, we present baseline results using various state-of-the-art pre-trained models: encoder-only and sequence-to-sequence; and conduct quantitative and qualitative error analysis. We found that pre-trained multilingual models generally outperform monolingual models on reference text and ASR output and encoders outperform sequence-to-sequence models in NER tasks. By translating the transcripts, the dataset can also be utilised for text NER in the medical domain in other languages than Vietnamese. All code, data and models are publicly available: https://github.com/leduckhai/MultiMed/tree/master/VietMed-NER.
Multilingual Large Language Models (LLMs) have gained large popularity among Natural Language Processing (NLP) researchers and practitioners. These models, trained on huge datasets, show proficiency across various languages and demonstrate effectiveness in numerous downstream tasks. This paper navigates the landscape of multilingual LLMs, providing an introductory overview of their technical aspects. It explains underlying architectures, objective functions, pre-training data sources, and tokenization methods. This work explores the unique features of different model types: encoder-only (mBERT, XLM-R), decoder-only (XGLM, PALM, BLOOM, GPT-3), and encoder-decoder models (mT5, mBART). Additionally, it addresses one of the significant limitations of multilingual LLMs - the curse of multilinguality - and discusses current attempts to overcome it.
Despite their remarkable ability to capture linguistic nuances across diverse languages, questions persist regarding the degree of alignment between languages in multilingual embeddings. Drawing inspiration from research on high-dimensional representations in neural language models, we employ clustering to uncover latent concepts within multilingual models. Our analysis focuses on quantifying the \textit{alignment} and \textit{overlap} of these concepts across various languages within the latent space. To this end, we introduce two metrics \CA{} and \CO{} aimed at quantifying these aspects, enabling a deeper exploration of multilingual embeddings. Our study encompasses three multilingual models (\texttt{mT5}, \texttt{mBERT}, and \texttt{XLM-R}) and three downstream tasks (Machine Translation, Named Entity Recognition, and Sentiment Analysis). Key findings from our analysis include: i) deeper layers in the network demonstrate increased cross-lingual \textit{alignment} due to the presence of language-agnostic concepts, ii) fine-tuning of the models enhances \textit{alignment} within the latent space, and iii) such task-specific calibration helps in explaining the emergence of zero-shot capabilities in the models.\footnote{The code is available at \url{https://github.com/baselmousi/multilingual-latent-concepts}}