Papers by Tomoyuki Kajiwara
Language-agnostic Representation from Multilingual Sentence Encoders for Cross-lingual Similarity Estimation (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to extract language-specific information from multilingual sentence embeddings are remarkably successful in cross-lingual and multilingual NLU tasks. |
| Approach: | They propose to extract language-specific information from the original embedding and use it to retrieve an embeddable that fully represents the sentence’s meaning. |
| Outcome: | The proposed method outperforms baselines on cross-lingual sentences even in low-resource language pairs where only tens of thousands of parallel sentence pairs are available. |
Text Classification with Negative Supervision (2020.acl-main)
Copied to clipboard
| Challenge: | Existing models for text representations have shown state-of-the-art performance on text classification tasks, however, the discrepancy between semantic similarity of texts and labelling standards affects classifiers. |
| Approach: | They propose a simple multitask learning model that uses negative supervision to generate distinct representations for texts with different labels. |
| Outcome: | The proposed model outperforms state-of-the-art models on classification tasks in three different languages. |
A Benchmark Dataset for Multi-Level Complexity-Controllable Machine Translation (2022.lrec-1)
Copied to clipboard
Kazuki Tani, Ryoya Yuasa, Kazuki Takikawa, Akihiro Tamura, Tomoyuki Kajiwara, Takashi Ninomiya, Tsuneo Kato
| Challenge: | Existing test datasets for MLCC-MT have three problems: A source language sentence and its simplified target language sentence are not necessarily exactly parallel. |
| Approach: | They propose to use a test dataset to evaluate multi-level complexity-controllable machine translation (MLCC-MT) their results are compared to a standard test dataset constructed from the Newsela corpus . |
| Outcome: | The proposed test dataset is based on the Newsela corpus and is released . it includes automatic filtering, manual check for parallel target language sentences . |
Disentangling Meaning and Language Components in Diverse Multilingual Sentence Embeddings (2026.acl-srw)
Copied to clipboard
| Challenge: | Existing studies have reported language specificity in multilingual sentence embeddings, resulting in language-specific subspaces. |
| Approach: | They propose to disentangle multilingual sentence embeddings into language-dependent and language-agnostic components to improve cross-lingual similarity estimation. |
| Outcome: | The proposed methods improve cross-lingual similarity estimation across multiple embeddings. |
Adversarial Training on Disentangling Meaning and Language Representations for Unsupervised Quality Estimation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods for unsupervised quality estimation of machine translation are limited to several major language pairs. |
| Approach: | They propose a method to distill language-agnostic meaning embeddings from multilingual sentence encoders for unsupervised quality estimation of machine translation. |
| Outcome: | The proposed method achieves higher correlations with human evaluations on unsupervised translation quality estimation. |
DIRECT: Direct and Indirect Responses in Conversational Text Corpus (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Neural conversation models have been able to generate fluent responses through training on a dialogue corpus, but they lack the ability to reveal the implied intentions of users. |
| Approach: | They propose to train neural conversation models on a dialogue corpus that provides pragmatic paraphrases to advance techniques for natural language understanding in dialogue systems. |
| Outcome: | The proposed corpus provides 71,498 pairs of indirect–direct utterance pairs accompanied by a multi-turn dialogue history extracted from the MultiWoZ dataset. |
Transfer Fine-tuning for Quality Estimation of Text Simplification (2024.lrec-main)
Copied to clipboard
| Challenge: | Experimental results show that quality estimation of text simplification models can be improved on a small labeled corpus. |
| Approach: | They propose a method to train quality estimation of text simplification on a small-scale labeled corpus prior to fine-tuning pre-trained language models. |
| Outcome: | The proposed method improves quality estimation of text simplification on a small-scale labeled corpus. |
Utilizing Longer Context than Speech Bubbles in Automated Manga Translation (2024.lrec-main)
Copied to clipboard
Hiroto Kaino, Soichiro Sugihara, Tomoyuki Kajiwara, Takashi Ninomiya, Joshua B. Tanner, Shonosuke Ishiwatari
| Challenge: | Existing methods to capture contextual information for manga machine translation are difficult to perform . unofficially translated pirated copies of manga are circulating overseas in large numbers . |
| Approach: | They propose two new ways to capture broader contextual information in manga machine translation . scene-based translation considers previous scene and broader context information . detailed analysis reveals the effect of zero-anaphora resolution in translation - highlighting the usefulness of longer contextual information if manga is translated in Japanese . |
| Outcome: | The proposed methods improve translation quality for manga (Japanese-style comics) the results show that the combined methods achieve the highest quality. |
Controllable Text Simplification with Deep Reinforcement Learning (2022.aacl-short)
Copied to clipboard
| Challenge: | Existing methods for controlling sentence difficulty have not taken into account sentence-level difficulties. |
| Approach: | They propose a method for controlling the difficulty of a sentence based on deep reinforcement learning. |
| Outcome: | The proposed method generates sentences of appropriate difficulty for the target audience through reinforcement learning. |
Controllable Paraphrase Generation for Semantic and Lexical Similarities (2024.lrec-main)
Copied to clipboard
| Challenge: | Lexically diverse paraphrases are crucial in data augmentation because they enhance the linguistic diversity of the corpus. |
| Approach: | They propose a controllable model for semantic and lexical similarities by attaching tags to the head of the input sentence. |
| Outcome: | The proposed model can paraphrase an input sentence according to the tags specified. |
A Japanese Dataset for Subjective and Objective Sentiment Polarity Classification in Micro Blog Domain (2022.lrec-1)
Copied to clipboard
Haruya Suzuki, Yuto Miyauchi, Kazuki Akiyama, Tomoyuki Kajiwara, Takashi Ninomiya, Noriko Takemura, Yuta Nakashima, Hajime Nagahara
| Challenge: | Existing studies on emotion analysis have studied the analysis of basic emotions and sentiment polarity independently. |
| Approach: | They extend the WRIME dataset with basic emotion intensity from both the writer's subjective and reader's perspective to include the Japanese sentiment polarity. |
| Outcome: | The proposed dataset is the first large-scale corpus to annotate both basic emotions and sentiment polarity labels from both the writer’s and reader’s perspectives. |
Definition Modelling for Appropriate Specificity (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing definition generation techniques have faced various problems such as the out-of-vocabulary problem and over/under-specificity problems. |
| Approach: | They propose to leverage a pre-trained encoder-decoder model and introduce a re-ranking mechanism to model specificity in definitions. |
| Outcome: | The proposed method significantly outperforms the state-of-the-art method on standard evaluation datasets and shows that it addresses the over/under-specificity problems. |
Contextualized context2vec (D19-55)
Copied to clipboard
| Challenge: | Lexical substitution ranks substitution candidates from the viewpoint of paraphrasability for a target word in a given sentence. |
| Approach: | They propose a method that combines two approaches to contextualize word embeddings for lexical substitution. |
| Outcome: | The proposed method outperforms the current state-of-the-art method and assigns English proficiency levels to all target words and substitution candidates. |
Evaluation Dataset for Japanese Medical Text Simplification (2024.naacl-srw)
Copied to clipboard
| Challenge: | Existing studies on medical text simplification in English have not been well explored in Japanese because of the lack of a parallel corpus of this domain. |
| Approach: | They propose a lexically constrained reranking method that allows to avoid technical terms to be output. |
| Outcome: | The proposed method improves on the weblogs of Japanese patients and reduces the need for a training corpus. |
WRIME: A New Dataset for Emotional Intensity Estimation with Subjective and Objective Annotations (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing studies on emotion analysis use subjective emotional intensity labels by the writers and objective ones by the readers. |
| Approach: | They annotate 17,000 SNS posts with both the writer's subjective emotional intensity and the reader's objective emotional intensity to construct a Japanese emotion analysis dataset. |
| Outcome: | The results show that the reader cannot fully detect the emotions of the writer, especially anger and trust. |
Domain Adaptation of Image Encoder for Multimodal Manga Translation (2026.eacl-srw)
Copied to clipboard
| Challenge: | Existing machine translation systems lack sufficient manga comprehension capabilities when utilizing image information. |
| Approach: | They propose a domain-adapted image encoder training method for manga . the method trains encoders to acquire visual features that consider the structural and sequential characteristics of the manga based on a Japanese-English translation task. |
| Outcome: | The proposed method improves translation evaluation metrics in Japanese-English translation task compared to the conventional method . |
Metric for Automatic Machine Translation Evaluation based on Universal Sentence Representations (N18-4)
Copied to clipboard
| Challenge: | Sentence representations can capture information that cannot be captured by local features based on character or word Ngrams. |
| Approach: | They propose a supervised regression model using universal sentence representations capable of capturing information that cannot be captured by local features based on character or word Ngrams. |
| Outcome: | The proposed model achieves state-of-the-art performance with only sentence representation features . |
Edit-Constrained Decoding for Sentence Simplification (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies have shown that lexically constrained decoding is effective for sentence simplification, but their constraints can be loose and may lead to sub-optimal generation. |
| Approach: | They propose an edit operation based on lexically constrained decoding for sentence simplification using a dictionary of technical terms as constraints. |
| Outcome: | The proposed method outperforms previous studies on English simplification corpora and is based on lexical paraphrasing. |
Multimodal Neural Machine Translation Using Synthetic Images Transformed by Latent Diffusion Model (2023.acl-srw)
Copied to clipboard
| Challenge: | Existing methods to translate source language sentences using images are not optimal for machine translation. |
| Approach: | They propose a new multimodal neural machine translation model using synthetic images transformed by a latent diffusion model. |
| Outcome: | The proposed model improves translation performance on English-German translation tasks using the Multi30k dataset. |
Negative Lexically Constrained Decoding for Paraphrase Generation (P19-1)
Copied to clipboard
| Challenge: | Paraphrase generation is a monolingual machine translation problem. |
| Approach: | They propose a neural model that first identifies words in the source sentence that should be paraphrased and then decodes them by negative lexical constraints. |
| Outcome: | The proposed model improves paraphrase generation by making necessary rewrites to an input sentence. |
Edit Distance Based Curriculum Learning for Paraphrase Generation (2021.acl-srw)
Copied to clipboard
| Challenge: | Existing studies show that curriculum learning improves translation quality on machine translation . paraphrase generation allows a certain level of semantic divergence between source and target . |
| Approach: | They propose to apply curriculum learning to paraphrase generation for the first time . they propose to use edit distance to improve paraphrase quality . |
| Outcome: | The proposed method improves paraphrase generation quality, compared with previous methods . it uses edit distance, which is not possible for previous methods, the authors say . |
Annotation of Adverse Drug Reactions in Patients’ Weblogs (2020.lrec-1)
Copied to clipboard
| Challenge: | Adverse drug reactions are a severe problem that significantly degrade quality of life and make the therapeutic approach unacceptable. |
| Approach: | They crawled patient’s weblog articles shared on an online patient-networking platform and annotated the effects of drugs therein reported. |
| Outcome: | The proposed dataset is unique for the richness of annotated information, including detailed descriptions of drug reactions with full context. |
Text Simplification with Reinforcement Learning Using Supervised Rewards on Grammaticality, Meaning Preservation, and Simplicity (2020.aacl-srw)
Copied to clipboard
| Challenge: | Existing studies in text-to-text generation do not align with human-perspectives for these perspectives. |
| Approach: | They propose to use BERT regressors fine-tuned for grammaticality, meaning preservation, and simplicity as reward estimators to optimize rewards for text simplification. |
| Outcome: | The proposed method achieves text simplification conforming to human-perspectives. |
SOME: Reference-less Sub-Metrics Optimized for Manual Evaluations of Grammatical Error Correction (2020.coling-main)
Copied to clipboard
| Challenge: | Existing reference-less metrics are not optimized for manual evaluations of system outputs because no dataset exists for manual analysis. |
| Approach: | They propose a reference-less metric trained on manual evaluations of system outputs for grammatical error correction. |
| Outcome: | The proposed metric improves correlation with manual evaluation in system- and sentence-level meta-evaluation. |
Distinct Label Representations for Few-Shot Text Classification (2021.acl-short)
Copied to clipboard
| Challenge: | Existing methods for few-shot text classification ignore the semantic relevance of labels and are difficult to train because of the lack of training examples. |
| Approach: | They propose a method that generates distinct label representations that embed information specific to each label. |
| Outcome: | The proposed method significantly improves few-shot text classification across models and datasets. |
Distilling Word Meaning in Context from Pre-trained Language Models (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to transform contextualised representations weaken excessive effects of contextual information. |
| Approach: | They propose a self-supervised learning method that distils word meaning in context from a pre-trained masked language model. |
| Outcome: | The proposed method outperforms the state-of-the-art method for lexical semantics and STS estimation. |
Distractor Generation for Fill-in-the-Blank Exercises by Question Type (2023.acl-srw)
Copied to clipboard
| Challenge: | Existing studies have generated words that are semantically similar to the correct words as distractors for fill-in-the-blank questions. |
| Approach: | They propose a method to automatically generate distractors for fill-in-the-blank questions in entrance examinations for Japanese universities. |
| Outcome: | The proposed method is effective on 500 actual questions on English fill-in-the-blank questions in Japanese universities. |
Emotional Intensity Estimation based on Writer’s Personality (2022.aacl-srw)
Copied to clipboard
| Challenge: | Existing emotion analysis models are difficult to accurately estimate the writer’s subjective emotions behind the text. |
| Approach: | They propose a method for personalized emotional intensity estimation based on a writer's personality test for Japanese SNS posts. |
| Outcome: | The proposed method improves on the existing method and the proposed hybrid model achieved state-of-the-art performance. |
JADE: Corpus for Japanese Definition Modelling (2022.lrec-1)
Copied to clipboard
| Challenge: | Existing corpus for definition modelling techniques is limited to English . this study aimed to develop a corpus that provides definitions of words and phrases . |
| Approach: | They investigated and released a corpus for Japanese definition modelling . the JADE provides 630k sets of targets, their definitions, and usage examples as contexts . |
| Outcome: | The JADE corpus provides 630k sets of targets, their definitions, and usage examples as contexts for 41k unique targets. |
Multi-Source Text Classification for Multilingual Sentence Encoder with Machine Translation (2024.naacl-srw)
Copied to clipboard
| Challenge: | Pre-trained multilingual sentence encoders suffer from performance degradation for non-English languages. |
| Approach: | They propose a method of machine translating a source sentence into English and then inputting it together with the source sentence in a multi-source manner. |
| Outcome: | The proposed method improves the performance of pre-trained multilingual sentence encoders in Japanese on sentiment analysis and topic classification tasks. |
Tiny Word Embeddings Using Globally Informed Reconstruction (2020.coling-main)
Copied to clipboard
| Challenge: | Existing methods for word embedding reconstruction use only local information of subwords and pre-trained word embeds. |
| Approach: | They propose a global loss function that uses words other than the target word to improve word embedding reconstruction by a factor of 200. |
| Outcome: | The proposed method reduces the model size of pre-trained word embeddings by a factor of 200 while preserving its quality. |
MultiMSD: A Corpus for Multilingual Medical Text Simplification from Online Medical References (2025.findings-acl)
Copied to clipboard
| Challenge: | Medical texts contain technical terms, and non-experts often cannot use information effectively. |
| Approach: | They propose a method for training medical text simplification models to actively paraphrase medical terms. |
| Outcome: | The proposed method improves the performance of medical text simplification in nine languages. |
Word Complexity Estimation for Japanese Lexical Simplification (2020.lrec-1)
Copied to clipboard
| Challenge: | Experimental results show that the proposed method achieves the highest performance of Japanese lexical simplification. |
| Approach: | They propose a large-scale word complexity lexicon, a synonym lexicone and a toolkit for developing and benchmarking Japanese lexical simplification systems. |
| Outcome: | The proposed method achieves the highest performance of Japanese lexical simplification. |
Automatic Decomposition of Text Editing Examples into Primitive Edit Operations: Toward Analytic Evaluation of Editing Systems (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods to automate text editing tasks are blackboxed and do not understand the behavior of the systems. |
| Approach: | They propose a task of automatic decomposition of text editing examples into primitive edit operations by using a phrase aligner and a large language model. |
| Outcome: | The proposed method perfectly decomposes 44% and 64% of editing examples . Detailed analyses also provide insights into the difficulties of this task . |
Controllable Text Simplification with Lexical Constraint Loss (P19-2)
Copied to clipboard
| Challenge: | Existing models that only consider the sentence level generate words beyond the target level. |
| Approach: | They propose a method to control the level of a sentence in a text simplification task . they add the target grade level as input and weight words in the loss function . |
| Outcome: | The proposed method improves both BLEU and SARI scores and achieves aggressive rewriting. |
Paraphrase-based Contrastive Learning for Sentence Pair Modeling (2025.naacl-srw)
Copied to clipboard
| Challenge: | Existing methods to improve performance of sentence pair modeling are not available on a large-scale for non-English languages. |
| Approach: | They propose a method to apply contrastive learning to pre-trained masked language models . they use sentence embeddings of paraphrase pairs to make similar sentences . |
| Outcome: | The proposed method can be used on four sentence pair modeling tasks in English and Japanese. |
Self-Ensemble of N-best Generation Hypotheses by Lexically Constrained Decoding (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies have improved generation quality by explicitly reranking N-best candidates. |
| Approach: | They propose a method that ensembles N-best hypotheses to improve natural language generation by combining high-quality fragments of N- best hypothese . they use tokens that should or should not be present in the final output as lexical constraints to improve quality of generation. |
| Outcome: | Empirical results show that the proposed method outperforms strong N-best reranking methods on paraphrase generation, summarisation, and constrained text generation. |
CEFR-Based Sentence Difficulty Annotation and Assessment (2022.emnlp-main)
Copied to clipboard
| Challenge: | Controllable text simplification is a crucial assistive technique for language learning and teaching. |
| Approach: | They propose a sentence-level assessment model to handle unbalanced level distribution . previous studies have suggested that controllable text simplification is difficult to apply . |
| Outcome: | The proposed method outperforms baselines in readability assessment by scoring macro-F1 on the level assessment. |
SAPPHIRE: Simple Aligner for Phrasal Paraphrase with Hierarchical Representation (2020.lrec-1)
Copied to clipboard
| Challenge: | Monolingual phrase alignment is a fundamental problem in natural language understanding and crucial technique in various applications. |
| Approach: | They propose a simple Aligner for Phrasal Paraphrase with HIerarchical REpresentation that uses word embeddings to train phrase alignments. |
| Outcome: | The proposed algorithm outperforms the previous methods and establishes the state-of-the-art. |