Papers with Thai
Misspelling Semantics in Thai (2022.lrec-1)
Copied to clipboard
| Challenge: | In English, more than 70% of documents on the internet contain some form of misspelling . misspellers can be used as prosody to provide additional clues about the writer's attitude . |
| Approach: | They propose two ways to incorporate misspelling semantics into user-generated content . they propose a method to boost micro F1 score by 0.4-2% . |
| Outcome: | The proposed methods can boost the micro F1 score up to 0.4-2% while normalising misspelling is harmful and suboptimal. |
Robust Neural Machine Translation for Abugidas by Glyph Perturbation (2024.eacl-short)
Copied to clipboard
| Challenge: | Neural machine translation systems are vulnerable when trained on limited data. |
| Approach: | They propose to add noise to the training phase to increase robustness of NMT systems trained on limited data. |
| Outcome: | The proposed training strategy overcomes noise and improves robustness for low-resource tasks for abugida glyphs. |
SeaLLMs - Large Language Models for Southeast Asia (2024.acl-demos)
Copied to clipboard
Xuan-Phi Nguyen, Wenxuan Zhang, Xin Li, Mahani Aljunied, Zhiqiang Hu, Chenhui Shen, Yew Ken Chia, Xingxuan Li, Jianyu Wang, Qingyu Tan, Liying Cheng, Guanzheng Chen, Yue Deng, Sen Yang, Chaoqun Liu, Hang Zhang, Lidong Bing
| Challenge: | Existing large language models favor high-resource languages, such as English, at the expense of low-resourced and regional languages. |
| Approach: | They propose a series of language models that specifically focuses on Southeast Asian languages. |
| Outcome: | SeaLLM models outperform ChatGPT-3.5 in non-Latin languages by large margins . linguistic disparity impedes access to state-of-the-art AI technologies for non-English-speaking populations . |
Seed-Free Synthetic Data Generation Framework for Instruction-Tuning LLMs: A Case Study in Thai (2024.acl-srw)
Copied to clipboard
| Challenge: | Xue et al., 2024) have demonstrated that large language models can perform at human level across multitudes of tasks and domains. |
| Approach: | They propose a seed-free framework for generating synthetic instruction-tuning data that incorporates fluency, diversity, and cultural context. |
| Outcome: | The proposed framework achieves competitive performance using only 5,000 instructions compared to state-of-the-art Thai LLMs trained on hundreds of thousands of instructions. |
Scaling Under-Resourced TTS: A Data-Optimized Framework with Advanced Acoustic Modeling for Thai (2025.acl-industry)
Copied to clipboard
| Challenge: | Text-to-speech (TTS) systems are limited by limited data and linguistic complexities. |
| Approach: | They propose a data-optimized framework with an advanced acoustic model to build high-quality TTS systems for low-resource scenarios. |
| Outcome: | The proposed framework enables zero-shot voice cloning and improved performance across diverse client applications, including finance, healthcare, education, and law. |
Sailor: Open Language Models for South-East Asia (2024.emnlp-demo)
Copied to clipboard
| Challenge: | Large language models (LLMs) rely on English data for training, but are often not comparable across other languages. |
| Approach: | They propose to develop a family of open language models for SEA languages . they use BPE dropout, aggressive data cleaning and deduplication to improve model robustness . |
| Outcome: | The proposed models perform well across four benchmarks, including commonsense reasoning, question answering, reading comprehension and examination. |
An Information-Theoretic Approach and Dataset for Probing Gender Stereotypes in Multilingual Masked Language Models (2022.findings-naacl)
Copied to clipboard
| Challenge: | Pretrained language models (PLMs) have been shown to encapsulate social biases, including those relating to gender and race. |
| Approach: | They propose a new bias measure based on Jensen–Shannon divergence that retains more information from the model output probabilities than other previously proposed bias measures. |
| Outcome: | The proposed measure outperforms CrowS-Pairs and other similar measures for non-English datasets. |
Simplified Abugidas (P18-2)
Copied to clipboard
| Challenge: | Abugidas are writing systems where consonant letters represent syllables with a default vowel and other vowels are denoted by diacritics. |
| Approach: | They investigated the feasibility of recovering the original text written in an abugida after omitting subordinate diacritics and merging consonant letters with similar phonetic values. |
| Outcome: | The proposed method recovers the original text written in an abugida with 94% - 97% accuracy at the top-1 level and 98% - 99% at the bottom-4 level even after omitting most diacritics and merging the remaining 30 - 50 characters into 21 graphemes. |
Handling Cross- and Out-of-Domain Samples in Thai Word Segmentation (2021.findings-acl)
Copied to clipboard
Peerat Limkonchotiwat, Wannaphong Phatthiyaphaibun, Raheem Sarwar, Ekapol Chuangsuwanich, Sarana Nutanong
| Challenge: | Word segmentation is domain-dependent, which can be a challenge in low-resource languages like Thai and Urdu . a framework to handle out-of-domain inputs is proposed to improve word segmentation . |
| Approach: | They propose a domaingeneric domain adaptation framework and data augmentation technique to combat low-resource problems. |
| Outcome: | The proposed model outperforms the state-of-the-art Thai word segmentation method in out-of domain scenarios. |
FiNERweb: Datasets and Artifacts for Scalable Multilingual Named Entity Recognition (2026.findings-eacl)
Copied to clipboard
| Challenge: | Named entity recognition (NER) is the task of identifying tokens that belong to a predefined set of classes such as "person" or "location" |
| Approach: | They propose a dataset-creation pipeline that scales the teacher-student paradigm to 91 languages and 25 scripts. |
| Outcome: | The proposed model achieves comparable or improved performance in English, Thai, and Swahili despite being trained on 19x less data than strong baselines. |
GigaSpeech 2: An Evolving, Large-Scale and Multi-domain ASR Corpus for Low-Resource Languages with Automated Crawling, Transcription and Refinement (2025.acl-long)
Copied to clipboard
Yifan Yang, Zheshu Song, Jianheng Zhuo, Mingyu Cui, Jinpeng Li, Bo Yang, Yexing Du, Ziyang Ma, Xunying Liu, Ziyuan Wang, Ke Li, Shuai Fan, Kai Yu, Wei-Qiang Zhang, Guoguo Chen, Xie Chen
| Challenge: | GigaSpeech 2 is a large-scale, multi-domain, multilingual speech recognition corpus for low-resource languages. |
| Approach: | They propose a large-scale, multi-domain, multilingual speech recognition corpus for low-resource languages and an automated pipeline for data crawling, transcription, and label refinement. |
| Outcome: | The proposed corpus reduces the word error rate for Thai, Indonesian, and Vietnamese on a realistic YouTube test set by 25% to 40% compared to Whisper large-v3. |
WangchanThaiInstruct: An instruction-following Dataset for Culture-Aware, Multitask, and Multi-domain Evaluation in Thai (2025.emnlp-main)
Copied to clipboard
Peerat Limkonchotiwat, Pume Tuchinda, Lalita Lowphansirikul, Surapon Nonesung, Panuthep Tasawong, Alham Fikri Aji, Can Udomcharoenchaikit, Sarana Nutanong
| Challenge: | Existing benchmarks for large language models rely on translations, missing cultural and domain specificity. |
| Approach: | They present a human-authored dataset for evaluation and instruction tuning in Thai . findings highlight need for culturally and professionally grounded instruction data . |
| Outcome: | a human-authored dataset for evaluation and instruction tuning in Thai outperforms translation-based models . findings highlight need for culturally and professionally grounded instruction data . |
More Than Words: Collocation Retokenization for Latent Dirichlet Allocation Models (2022.findings-acl)
Copied to clipboard
| Challenge: | Latent Dirichlet Allocation models ingest words to discover their latent topics . but it is unclear how to achieve the best results for languages without marked word boundaries . |
| Approach: | They propose to use retokenization to merge frequent token ngrams into collocations in input to a Latent Dirichlet Allocation model. |
| Outcome: | The proposed model improves topic coherence and coherency in Chinese and Thai . the proposed model is more coherent and clearer than unmerged models . |
Can LLMs Simulate L2-English Dialogue? An Information-Theoretic Analysis of L1-Dependent Biases (2025.acl-long)
Copied to clipboard
Rena Wei Gao, Xuetong Wu, Tatsuki Kuribayashi, Mingrui Ye, Siya Qi, Carsten Roever, Yuanxing Liu, Zheng Yuan, Jey Han Lau
| Challenge: | Large Language Models (LLMs) can simulate non-native-like English use observed in human second language (L2) learners interfered with by their native first language (N1) knowledge. |
| Approach: | They use large language models to simulate non-native-like English use observed in human second language (L2) learners, and then compare their outputs to real L2 learner data. |
| Outcome: | The proposed models replicate L1-dependent patterns observed in human second language (L2) learners, with distinct influences from various languages. |
Representing the Under-Represented: Cultural and Core Capability Benchmarks for Developing Thai Large Language Models (2025.coling-main)
Copied to clipboard
| Challenge: | Rapid advancements in large language models have highlighted the need for robust evaluation frameworks that assess their core capabilities. |
| Approach: | They propose two benchmarks to assess core capabilities of large language models . current benchmarks for Thai focus mainly on traditional NLP tasks . |
| Outcome: | The proposed benchmarks are based on evaluations of various LLMs with multi-lingual capabilities and are publicly available to encourage further research and development for Thai LLM. |
Grapheme-to-Phoneme Conversion for Thai using Neural Regression Models (2022.naacl-main)
Copied to clipboard
| Challenge: | Grapheme-to-phoneme conversion is a task of converting grapheme sequences into phoneme sequence. |
| Approach: | They propose a Thai grapheme-to-phoneme conversion method that uses neural networks to predict the similarity between a candidate and the correct pronunciation. |
| Outcome: | The proposed method can be applied to other languages than Thai . it is comparable to encoder-decoder models in accuracy and accuracy, it shows . |
Cross-Lingual NLU: Mitigating Language-Specific Impact in Embeddings Leveraging Adversarial Learning (2024.lrec-main)
Copied to clipboard
| Challenge: | Low-resource languages and computational expenses pose significant challenges in the domain of large language models. |
| Approach: | They propose a novel approach that uses adversarial techniques to mitigate the impact of language-specific information in contextual embeddings generated by large multilingual language models. |
| Outcome: | The proposed approach excels in zero-shot scenarios for Latin languages like Spanish, but fails to perform for languages distant from English, such as Thai and Persian. |
Cross-lingual Transfer Learning for Multilingual Task Oriented Dialog (N19-1)
Copied to clipboard
| Challenge: | a lack of multilingual training data has hindered development of conversational AI models for task-oriented tasks . a new data set of 57k annotated utterances in english, spanish, and Thai is used to evaluate cross-lingual methods . |
| Approach: | They present a data set of 57k annotated utterances in English, Spanish and Thai . they evaluate three different cross-lingual transfer methods to identify user intents and slots . |
| Outcome: | The proposed model outperforms existing methods in English, Spanish and Thai . the proposed model is based on training data from three languages . |
Improving Low-Resource Named Entity Recognition using Joint Sentence and Token Labeling (2020.acl-main)
Copied to clipboard
| Challenge: | Existing models for named entity recognition (NER) use sentence-level labels, which are expensive to obtain, to improve NER. |
| Approach: | They propose a sentence-level named entity recognition model that uses sentence-based labels that are easy to obtain. |
| Outcome: | The proposed model produces 3.78%, 4.20%, 2.08% improvements in F1 over the baseline on e-commerce product titles in Vietnamese, Thai, and Indonesian, respectively. |
A Simple and Effective Approach to Robust Unsupervised Bilingual Dictionary Induction (2020.coling-main)
Copied to clipboard
| Challenge: | Recent work has questioned the robustness of unsupervised bilingual dictionary induction methods on distant language pairs. |
| Approach: | They propose an iterative dimension reduction method to bridge this gap . they propose a method that initializes and self-learning and inducing a dictionary . |
| Outcome: | The proposed method achieves 13.64 55.53% accuracy between English and four distant languages. |
Overcoming Catastrophic Forgetting in Zero-Shot Cross-Lingual Generation (2022.emnlp-main)
Copied to clipboard
| Challenge: | generative multilingual models fine-tuned on English forget to generate non-English data when labeled data is only available in English . generative models fine tuned on English fail to generate multilingual summarization tasks when labeling data is available in other languages . |
| Approach: | They propose to use prompt tuning to overcome catastrophic forgetting in a generative task in . they assume a strict setting with no parallel data or machine translation . |
| Outcome: | The proposed method can overcome catastrophic forgetting to enable zero-shot cross-lingual generation. |
SEA-HELM: Southeast Asian Holistic Evaluation of Language Models (2025.findings-acl)
Copied to clipboard
Yosephine Susanto, Adithya Venkatadri Hulagadri, Jann Railey Montalan, Jian Gang Ngui, Xianbin Yong, Wei Qi Leong, Hamsawardhini Rengarajan, Peerat Limkonchotiwat, Yifan Mai, William Chandra Tjhi
| Challenge: | Existing LLM benchmarks are capable of evaluating specific capabilities in English as well as in various mid- to low-resource languages, but a comprehensive and culturally representative evaluation suite for the SEA languages has not been developed thus far. |
| Approach: | They propose a holistic linguistic and cultural LLM evaluation suite that emphasizes SEA languages and introduces a leaderboard that allows users to understand models’ multilingual and multicultural performance. |
| Outcome: | The proposed evaluation suite emphasizes SEA languages and supports Filipino, Indonesian, Tamil, Thai, and Vietnamese. |
Unlocking Multilingual Reasoning Capability of LLMs and LVLMs through Representation Engineering (2026.acl-long)
Copied to clipboard
Qiming Li, Xiaocheng Feng, Yixuan Ma, Ruihan Chen, Zihe Tong, Zekai Ye, Xiachong Feng, Libo Qin, Haoyu Ren, Kun Chen, Yunfei Lu, Dandan Tu, Bing Qin
| Challenge: | Existing approaches to enhance multilingual reasoning capabilities rely on costly multilingual training or employ prompting with external translation tools. |
| Approach: | They propose a training-free inference-time method to enhance multilingual reasoning capabilities via Representation Engineering without additional training data or tools. |
| Outcome: | The proposed method outperforms existing methods on four reasoning benchmarks in English and Thai and Swahili. |
When Meaning Travels: A Granular Lens on Hybrid-MoE’s Role in Idiomatic Understanding for Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | idioms provide a fascinating gateway to creativity, cultural values, historical context, and diverse perspectives inherent to diverse linguistic traditions. |
| Approach: | They propose a multimodal idiom corpus enriched with seven idiomatic tones . they propose idiomic hybridization framework that embeds multiple idiomatic expert opinions . |
| Outcome: | The proposed framework achieves 5–6% performance gains across advanced vision language models. |
NitiBench: Benchmarking LLM Frameworks on Thai Legal Question Answering Capabilities (2025.emnlp-main)
Copied to clipboard
Pawitsapak Akarajaradwong, Pirat Pothavorn, Chompakorn Chaksangchaichot, Panuthep Tasawong, Thitiwat Nopparatbundit, Keerakiat Pratai, Sarana Nutanong
| Challenge: | Large language models (LLMs) show promise in legal question answering (QA), yet Thai legal QA systems face challenges due to limited data and complex legal structures. |
| Approach: | They propose a benchmark which uses Thai financial laws and tax rulings to evaluate Thai legal QA systems. |
| Outcome: | The proposed benchmark compared retrieval-augmented generation and long-context LLM approaches across three key dimensions and found that they improve over naive methods. |