Papers by Yugo Murawaki
Latent Geographical Factors for Analyzing the Evolution of Dialects in Contact (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to analyze the evolution of dialects use admixture analysis, but such ancestral populations are hardly interpretable in the context of the tree model. |
| Approach: | They propose a probabilistic generative model that represents latent factors as geographical distributions and a tree model that can be alternatively represented as a set of geographical distribution. |
| Outcome: | The proposed model has higher affinity with the tree model because a tree can alternatively be represented as a set of geographical distributions. |
Principal Component Analysis as a Sanity Check for Bayesian Phylolinguistic Reconstruction (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods to reconstruct the evolutionary history of languages rely on the tree model . however, this assumption is violated to varying degrees due to contact and other factors . |
| Approach: | They propose a Bayesian tree model that assumes languages descended from a common ancestor and underwent modifications over time. |
| Outcome: | The proposed method visualizes anomalies in the form of jogging using synthetic and real data. |
A Knowledge-Augmented Neural Network Model for Implicit Discourse Relation Classification (C18-1)
Copied to clipboard
| Challenge: | Existing studies on implicit discourse relation classification have shown success using feedforward networks and convolutional neural networks. |
| Approach: | They propose to augment input text with external knowledge and context and adopt a neural network model that can effectively handle the augmented text. |
| Outcome: | The proposed model outperforms existing models on implicit discourse relation classification. |
Diversity-aware Event Prediction based on a Conditional Variational Autoencoder with Reconstruction (D19-60)
Copied to clipboard
| Challenge: | Typical event sequences are important class of commonsense knowledge . previous work in event prediction uses sequence-to-sequence models . however, what can happen after a given event is usually diverse . |
| Approach: | They propose to incorporate a conditional variational autoencoder into seq2seq for its ability to represent diverse next events as a probabilistic distribution. |
| Outcome: | The proposed model outperforms deterministic models in terms of precision and recall . the proposed model is based on a conditional variational autoencoder . |
Anchored Sliding Window: Toward Robust and Imperceptible Linguistic Steganography (2026.acl-long)
Copied to clipboard
| Challenge: | linguistic steganography assumes that stegographic texts are fragile to even minor modifications, compromising text quality. |
| Approach: | They propose an anchored sliding window framework to improve imperceptibility and robustness . they propose to include the prompt and a bridge context within the context window . |
| Outcome: | The proposed framework outperforms the baseline method in text quality, imperceptibility and robustness across diverse settings. |
Native-like Expression Identification by Contrasting Native and Proficient Second Language Speakers (2020.coling-main)
Copied to clipboard
| Challenge: | a novel task of native-like expression identification is proposed by contrasting texts written by native speakers and those by proficient second language speakers. |
| Approach: | They propose a task of native-like expression identification by contrasting texts written by native speakers and those by proficient second language speakers. |
| Outcome: | The proposed method uncovers linguistically interesting usages distinctive of native speech. |
Identifying Source Language Expressions for Pre-editing in Machine Translation (2024.lrec-main)
Copied to clipboard
| Challenge: | MT-mediated communication can benefit from pre-editing source language texts to ensure accurate transmission of intended meaning in the target language. |
| Approach: | They hypothesize that such expressions tend to be distinctive features of texts originally written in the source language rather than translations generated from the target language into the source languages. |
| Outcome: | The proposed method identified characteristic expressions of the native language despite the noise and inherent nuances of the task. |
What Language Do Non-English-Centric Large Language Models Think in? (2025.findings-acl)
Copied to clipboard
Chengzhi Zhong, Qianying Liu, Fei Cheng, Junfeng Jiang, Zhen Wan, Chenhui Chu, Yugo Murawaki, Sadao Kurohashi
| Challenge: | Despite their robust performance in English, these models often exhibit reduced proficiency in non-English languages, and their outputs may reflect an inherent bias toward English-centric perspectives. |
| Approach: | They categorize non-English-centric large language models into two groups: CPMs and BLMs, which are pre-trained on a balanced mix of multiple languages from scratch. |
| Outcome: | The proposed models exhibit a pronounced internal preference for English tokens when projected into the vocabulary space. |
CAPE: Context-Aware Personality Evaluation Framework for Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies use a context-free approach to assess humans . existing studies use the Disney World test, which ignores real-world applications . |
| Approach: | They propose a framework to assess personality traits in large language models . they use conversational history to quantify the consistency of LLM responses . |
| Outcome: | The proposed framework improves consistency of responses in large language models . it also shows that conversational history enhances consistency and personality shifts . |
Language Lives in Sparse Dimensions: Toward Interpretable and Efficient Multilingual Control for Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Prior studies show that large language models map multilingual content into English-aligned representations at intermediate layers before projecting them back into target-language token spaces in the later layers. |
| Approach: | They propose a method to identify and manipulate dimensions that are sparse and sparsity-based . they propose to use as few as 50 sentences of either parallel or monolingual data to manipulate these dimensions . |
| Outcome: | Experiments on a multilingual generation control task show the interpretability of these dimensions. |
Universal Dependencies Version 2 for Japanese (L18-1)
Copied to clipboard
Masayuki Asahara, Hiroshi Kanayama, Takaaki Tanaka, Yusuke Miyao, Sumire Uematsu, Shinsuke Mori, Yuji Matsumoto, Mai Omura, Yugo Murawaki
| Challenge: | UD Japanese resources are built on automatic conversion from several treebanks. |
| Approach: | They propose to port the word delimitation, POS, and syntactic relations of existing treebanks to UD Japanese . they discuss the issues of the UD scheme found through porting of the Japanese language . |
| Outcome: | The proposed UD Japanese resources are based on automatic conversion from treebanks. |
Improving Crowdsourcing-Based Annotation of Japanese Discourse Relations (L18-1)
Copied to clipboard
| Challenge: | Discourse parsing is an important task in natural language processing, but few languages have corpora annotated with discourse relations . crowdsourcing-based annotations are of poor quality and require expensive and time-consuming . et al. (2009) evaluated the quality of annotations using expert annotations. |
| Approach: | They construct a Japanese corpus with discourse annotations through crowdsourcing . they propose improvement techniques based on language tests . |
| Outcome: | The proposed methods improve the quality of the annotations, and will make them publicly available. |
Japanese Zero Anaphora Resolution Can Benefit from Parallel Texts Through Neural Transfer Learning (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Using a pretraining model, we find that the performance of Japanese zero anaphora resolution (ZAR) is improved by using machine translation. |
| Approach: | They propose to inject machine translation as an intermediate task between pretraining and ZAR by injecting machine translation into a pretrained BERT model and injecting it into MT. |
| Outcome: | The proposed framework shows that Japanese zero anaphora resolution (ZAR) can be improved by transfer learning from machine translation (MT). |
Building a Japanese Typo Dataset from Wikipedia’s Revision History (2020.acl-srw)
Copied to clipboard
| Challenge: | Typographical errors (typos) also occur in user generated content (UGC). |
| Approach: | They extract over half a million Japanese typo–correction pairs from Wikipedia’s revision history and combine character-based extraction rules, morphological analyzers to guess readings, and various filtering methods to address these challenges. |
| Outcome: | The proposed dataset extracts over half a million typo–correction pairs from Wikipedia’s revision history. |
Domain Transferable Semantic Frames for Expert Interview Dialogues (2024.lrec-main)
Copied to clipboard
| Challenge: | a dataset of interview dialogues with experts in the domains of culinary and gardening domains is used to structure domain-specific knowledge in expert interviews. |
| Approach: | They analyze interview dialogues with experts in the culinary and gardening domains to understand their domain knowledge structure. |
| Outcome: | The proposed framework is effective in eliciting critical skills in domains, the authors show . they use domain-agnostic labels to identify domain-specific knowledge in interviews . |
Frustratingly Easy Edit-based Linguistic Steganography with a Masked Language Model (2021.naacl-main)
Copied to clipboard
| Challenge: | linguistic steganography is the practice of concealing a secret message in some cover data such that an eavesdropper is not even aware of the existence of the secret message. |
| Approach: | They propose to use edit-based linguistic steganography to generate genuine-looking texts by using a masked language model that eliminates painstaking rule construction and has a high payload capacity. |
| Outcome: | The proposed method eliminates painstaking rule construction and has a high payload capacity for an edit-based model. |
Minimally Supervised Learning of Affective Events Using Discourse Relations (D19-1)
Copied to clipboard
| Challenge: | Existing methods for learning affective events that trigger positive or negative sentiment are difficult because of the unbounded combinatorial nature of language. |
| Approach: | They propose to propagate affective polarity using discourse relations using a small seed lexicon and large raw corpus. |
| Outcome: | The proposed method learns affective events effectively without manually labeled data, and improves supervised learning when labeles are small. |
Analyzing Correlated Evolution of Multiple Features Using Latent Representations (D18-1)
Copied to clipboard
| Challenge: | phylogenetic models allow quantitative analysis of evolution of a single categorical feature and a pair of binary features, but correlated evolution involving multiple discrete features is yet to be explored. |
| Approach: | They propose a latent representation-based analysis where discrete features are projected to a sequence of independent binary variables and phylogenetic inference is performed on the latent space. |
| Outcome: | The proposed model shows that languages sharing the same word order are not necessarily a coherent group but exhibit varying degrees of diachronic stability depending on other features. |
Addressing Tokenization Inconsistency in Steganography and Watermarking Based on Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models have improved the capacities and efficiency of text generation. |
| Approach: | They propose a method for tokenization inconsistency and a watermarking technique to address this problem. |
| Outcome: | The proposed methods improve fluency, imperceptibility, and anti-steganalysis capacity. |
Addressing Segmentation Ambiguity in Neural Linguistic Steganography (2022.aacl-short)
Copied to clipboard
| Challenge: | Recent studies on neural linguistic steganography ignore the fact that the sender must detokenize cover texts to avoid arousing the eavesdropper’s suspicion. |
| Approach: | They propose to decode a secret message in a way that does not arouse suspicion of the eavesdropper. |
| Outcome: | The proposed techniques are applicable to languages without explicit word boundaries. |
Adapting BERT to Implicit Discourse Relation Classification with a Focus on Discourse Connectives (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing studies on the performance of BERT for implicit discourse relation classification have not been conducted. |
| Approach: | They propose to apply BERT to implicit discourse relation classification by performing additional pre-training on text tailored to discourse relations. |
| Outcome: | The proposed methods outperform previous state-of-the-art models in many tasks. |
KWJA: A Unified Japanese Analyzer Based on Foundation Models (2023.acl-demo)
Copied to clipboard
Nobuhiro Ueda, Kazumasa Omura, Takashi Kodama, Hirokazu Kiyomaru, Yugo Murawaki, Daisuke Kawahara, Sadao Kurohashi
| Challenge: | KWJA supports a wide range of tasks including typo correction, word segmentation, word normalization, named entity recognition, dependency parsing, PAS analysis, bridging reference resolution, coreference resolution, and discourse relation analysis. |
| Approach: | They propose to build a Japanese text analyzer based on foundation models that performs a wide range of tasks. |
| Outcome: | The proposed model performs better in a multi-task manner than other analyzers with specialized models. |
Efficient Provably Secure Linguistic Steganography via Range Coding (2026.acl-long)
Copied to clipboard
| Challenge: | Linguistic steganography is a promising field in safeguarding information . previous methods have achieved perfect imperceptibility but at the expense of embedding capacity. |
| Approach: | They propose to use a classical entropy coding method to achieve secure steganography . they propose to employ a rotation mechanism to achieve embedding efficiency . |
| Outcome: | The proposed method outperforms existing methods in embedding capacity and embeddability. |
Annotating Modality Expressions and Event Factuality for a Japanese Chess Commentary Corpus (L18-1)
Copied to clipboard
| Challenge: | In recent years, there has been a surge of interest in the natural language processing related to the real world . shogi commentaries are an interesting testbed for these tasks, but can be grounded in the game tree . |
| Approach: | They propose to augment shogi commentaries with game states to generate a game commentary generator. |
| Outcome: | The proposed system can be used to ground symbols and events with factuality . it can be compared with other systems to find out if a commentator is a human . |
Persona Jailbreaking in Large Language Models (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing studies focus on narrative or role-playing tasks and overlook how adversarial conversational history alone can reshape induced personas. |
| Approach: | They propose a framework that embeds semantically loaded cues into user queries to gradually induce reverse personas. |
| Outcome: | The proposed framework predictably shifts personas, triggers collateral changes in correlated traits, and exhibits stronger effects in multi-turn settings. |
How Does Cognitive Bias Affect Large Language Models? A Case Study on the Anchoring Effect in Price Negotiation Simulations (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Cognitive biases can be observed in LLMs, affecting their reliability in real-world applications. |
| Approach: | They investigate the anchoring effect in LLM-driven price negotiations . reasoning models are less prone to the anchor effect, they find . |
| Outcome: | The proposed study shows that LLMs are influenced by the anchoring effect like humans . reasoning models are less prone to the anchor effect, but personality traits are not affected . |