Faculty Profiles - MATSUBARA, Shigeki

Capabilities and Challenges of LLMs in Metadata Extraction from Scholarly Papers

Other Link： https://link.springer.com/chapter/10.1007/978-981-96-0868-3_6

Watanabe Y., Ito K., Matsubara S.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 15493 LNCS page： 280 - 287 2025

Estimation of Relevance Between Datasets for Enhancing Accessibility of Research Artifacts

Ito K., Matsubara S.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 15494 LNCS page： 97 - 103 2025

The 30th Anniversary Initiatives of the Association for Natural Language Processing Open Access

Inoue Naoya, Ouchi Hiroki, Kawahara Daisuke, Kurohashi Sadao, Komachi Mamoru, Sudoh Katsuhito, Nagata Ryo, Matsubara Shigeki, Miyao Yusuke

Journal of Natural Language Processing Vol. 32 ( 1 ) page： 330 - 341 2025

Negation Scope Conversion for Unifying Negation-Annotated Datasets Open Access

Yoshida Asahi, Kato Yoshihide, Matsubara Shigeki

Journal of Natural Language Processing Vol. 32 ( 1 ) page： 300 - 329 2025

Language：English Publishing type：Research paper (scientific journal) Publisher：The Association for Natural Language Processing

Negation scope resolution is a technique that identifies the part of a sentence affected by the negation cue. The three major corpora used for it, the BioScope corpus, the SFU review corpus, and the Sherlock dataset, have different annotation schemes for negation scope. Due to the different annotations, it is difficult to use the three corpora together in the study of negation scope resolution. To address this issue by merging the corpora into a unified dataset based on a common annotation scheme, we propose a method for automatically converting the scopes of BioScope and SFU to those of Sherlock. We conducted an experiment to evaluate the accuracy of our method using a dataset obtained by manually annotating the negation scopes to a tiny portion of BioScope and SFU, verifying that our method can convert the scopes with high accuracy. In addition, we conducted another experiment to verify the effectiveness of our method from a pragmatic perspective, where we fine-tuned PLM-based negation scope resolution models using the unified dataset obtained by our method. The results demonstrated that the performances of the models increase when fine-tuned on the unified dataset, unlike the simply combined one, which supports the effectiveness of our method.

DOI： 10.5715/jnlp.32.300

Classification of URL Citations on Scholarly Papers Using Intermediate Task Training Open Access

WADA Kazuhiro, TSUNOKAKE Masaya, MATSUBARA Shigeki

IEICE Transactions on Information and Systems Vol. advpub ( 0 ) 2025

Language：English Publishing type：Research paper (scientific journal) Publisher：The Institute of Electronics, Information and Communication Engineers

Citations using URL (URL citations) that appear in scholarly papers can be used as an information source for the research resource search engines. In particular, the information about the types of cited resources and reasons for their citation is crucial to describe the resources and their relations in the search services. To obtain this information, previous studies proposed some methods for classifying URL citations. However, their methods trained the model using a simple fine-tuning strategy and exhibited insufficient performance. We propose a classification method using a novel intermediate task. Our method trains the model on our intermediate task of identifying whether sample pairs belong to the same class before being fine-tuned on the target task. In the experiment, our method outperformed previous methods using the simple fine-tuning strategy with higher macro F-scores for different model sizes and architectures. Our analysis results indicate that the model learns the class boundaries of the target task by training our intermediate task. Our intermediate task also demonstrated higher performance and computational efficiency than an alternative intermediate task using triplet loss. Finally, we applied our method to other text classification tasks and confirmed the effectiveness when a simple fine-tuning strategy does not stably work.

DOI： 10.1587/transinf.2024edp7149

Human Performance in Incremental Dependency Parsing: Dependency Structure Annotations and their Analyses Reviewed

Hiroki Unno, Tomohiro Ohno, Koichiro Ito, Shigeki Matsubara

Proceedings of the 37th Pacific Asia Conference on Language, Information and Computation (PACLIC-2024) 2024.12

Enhancing accessibility of research data through metadata conversion and registration in institutional repositories: Practices in the field of Solar-Terrestrial Physics Invited

NOSÉ Masahito, SHINBORI Atsuki, MIYOSHI Yoshizumi, HORI Tomoaki, OOHIRA Tsukasa, OKAMOTO Maiko, NAOE Chizuko, GAKIYA Rui, TANAKA Sachie, SAGARA Takeshi, AOKI Takaaki, MATSUBARA Shigeki, TAKAHASHI Ichiro, HAYASHI Hidekazu, YAMADA Kazunari, MINAMIYAMA Yasuyuki, TANAKA Yoshimasa, ABE Shuji, UENO Satoru, IMAJO Shun, SAITO Yasuo, ASHIKITA Takuya, HORI Yuko, SHIMIZU Toshiyuki, OKAMURA Nanako, HIRANO Kaoru

The Journal of Information Science and Technology Association Vol. 74 ( 11 ) page： 487 - 493 2024.11

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：Information Science and Technology Association, Japan

The “Inter-university Upper atmosphere Global Observation NETwork (IUGONET)” project has developed a metadata database for ground-based observation data in Solar-Terrestrial Physics. In order to make this metadata database accessible to a wide range of research communities and the general public, we converted metadata in different metadata schema and registered them in academic institutional repositories. As a result, through regular harvesting between databases, the metadata has also been registered in the Institutional Repositories Database (IRDB), the Data Catalog Cross-Search System, and Google’s Dataset Search, significantly enhancing the visibility and accessibility of research data. Furthermore, effective metadata management was implemented in response to research metadata maintenance requests from policy-making bodies.

DOI： 10.18919/jkg.74.11_487

Responsive Utterance Generation to Show Disagreement to Narrative in Attentive Listening Reviewed

Utilization of Text Data for Response Timing Detection in Attentive Listening Reviewed

Yu Watanabe, Koichiro Ito, Shigeki Matsubara

Proceedings of Interspeech-2024 page： 3565 - 3569 2024.9

On an Intermediate Task for Classifying URL Citations on Scholarly Papers Reviewed

Kazuhiro Wada, Masaya Tsunokake, Shigeki Matsubara

Proceeding of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING2024) 2024.5

Negation Scope Conversion: Towards a Unified Negation-Annotated Dataset Reviewed

Asahi Yoshida, Yoshihide Kato, Shigeki Matsubara

Proceeding of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING2024) 2024.5

Ito Koichiro, Murata Masaki, Ohno Tomohiro, Matsubara Shigeki

Journal of Natural Language Processing Vol. 31 ( 1 ) page： 212 - 249 2024.3

Authorship：Last author Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Association for Natural Language Processing

Conversational agents, such as communication robots, are expected to play a role in listening to narratives. To recognize these robots as the listeners, it is essential for them to have a function that indicates their attentive listening to a narrative. The basic response strategy in attentive listening involves generating responsive utterances that show acceptance. However, the narratives occasionally contain self-deprecation or modesty. In such cases, the listener must be able to produce a response that shows disagreement with the narrator's utterance, that is, a disagreement response. This study demonstrates the feasibility of generating appropriate disagreement responses. First, we define a method to tag the timing and expression of the disagreement response to narrative data in a non-real-time environment, and verify that the method enables to construct a corpus to which the disagreement response timing and response expression are comprehensively and stably assigned, respectively. Next, we implement detection and classification methods for disagreement response timing and expressions, respectively, based on a pre-trained transformer-based model and verify the feasibility of generating disagreement responses using the response corpus through experiments.

DOI： 10.5715/jnlp.31.212

Re-Evaluating Syntax-Based Negation Scope Resolution Open Access

YOSHIDA Asahi, KATO Yoshihide, MATSUBARA Shigeki

IEICE Transactions on Information and Systems Vol. E107.D ( 1 ) page： 165 - 168 2024.1

Language：English Publishing type：Research paper (scientific journal) Publisher：The Institute of Electronics, Information and Communication Engineers

Negation scope resolution is the process of detecting the negated part of a sentence. Unlike the syntax-based approach employed in previous researches, state-of-the-art methods performed better without the explicit use of syntactic structure. This work revisits the syntax-based approach and re-evaluates the effectiveness of syntactic structure in negation scope resolution. We replace the parser utilized in the prior works with state-of-the-art parsers and modify the syntax-based heuristic rules. The experimental results demonstrate that the simple modifications enhance the performance of the prior syntax-based method to the same level as state-of-the-art end-to-end neural-based methods.

DOI： 10.1587/transinf.2023edl8050

Web of Science

Enhancing Findability and Searchability of Research Data: Metadata Conversion and Registration in Institutional Repositories Open Access

Nosé M., Shinbori A., Miyoshi Y., Hori T., Ohira T., Hashiba J., Naoe C., Gakiya R., Okamoto M., Sagara T., Aoki T., Matsubara S., Takahashi I., Hayashi H., Yamada K., Minamiyama Y., Tanaka Y., Abe S., Ueno S., Imajo S., Saito Y., Ashikita T., Hori Y., Shimizu T., Okamura N., Hirano K., Bargatze L.

Data Science Journal Vol. 23 2024

Estimating Metadata of Research Artifacts to Enhance their Findability

Koichiro Ito, Shigeki Matsubara

Proceedings of 2024 IEEE 20th International Conference on e-Science (e-Science) page： 1 - 2 2024

Paper Recommendation Using Citation Contexts in Scholarly Documents Reviewed

Tomoki Ikoma, Shigeki Matsubara

Proceedings of the 37th Pacific Asia Conference on Language, Information and Computation (PACLIC 37) 2023.12

Automatic Insertion of Commas and Linefeeds into Lecture Transcripts Based on Multi-task Learning Reviewed

Zhicheng Fang, Masaki Murata, Shigeki Matsubara:

Proceedings of the 37th Pacific Asia Conference on Language, Information and Computation (PACLIC 37) 2023.12

On the Use of Language Models for Function Identification of Citations in Scholarly Papers Reviewed

Tomoki Ikoma and Shigeki Matsubara

Proceedings of the 2nd Workshop on Information Extraction from Scientific Publications (WIESP 2023) 2023.11

Revisiting Syntax-Based Approach in Negation Scope Resolution Reviewed

Asahi Yoshida, Yoshihide Kato and Shigeki Matsubara

Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*SEM2023) page： 18 - 23 2023.7

Natural Language Inference Using Neural Network and Tableau Method Reviewed

Saji Ayahito, Takao Daiki, Kato Yoshihide, Matsubara Shigeki

Journal of Natural Language Processing Vol. 30 ( 2 ) page： 632 - 663 2023.6

Authorship：Corresponding author Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Association for Natural Language Processing

Natural language inference (NLI) is the task of identifying the inferential relation between a text pair. In recent times, neural-based approaches have achieved high performance in NLI. However, they are unable to explain their reasoning processes. On the other hand, symbolic approaches have the advantage that their reasoning process is understandable to humans. This paper proposes a method for integrating a neural NLI model and the tableau proof system, with the latter explaining the reasoning processes. The standard tableau method decomposes logical formulas by applying inferential rules and checks for a valuation that satisfies the given constraints. Unlike the standard tableau method, the proposed method uses dependency structures as its components rather than logical formulas and employs a neural NLI model for the latter process. To analyze the behavior of our method, we conducted an experiment on the neural NLI model and the proposed method using SNLI corpus. In addition, we formalize our method model-theoretically and clarify the theoretical limitations of this method based on the formalization.

DOI： 10.5715/jnlp.30.632

Generation Timing Estimation of Responsive Utterances to Show Attentive Listening to Narrative Speech Reviewed

ITO Koichiro, MURATA Masaki, OHNO Tomohiro, MATSUBARA Shigeki

Vol. J106-A ( 3 ) page： 136 - 145 2023.3

Authorship：Last author Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Institute of Electronics, Information and Communication Engineers

Conversational agents, e.g., communication robots and smart speakers, are expected to listen to human narratives. For realizing such agents, it is necessary to generate responsive utterances at appropriate timings. This paper proposes a method to estimate the degree of appropriateness as generation timings of responsive utterances. The proposed method estimates the degree by encoding acoustic and linguistic narrative information with a transformer-based approach. To evaluate the performance, an experiment was conducted using 131,616 responsive utterances. The experimental results demonstrated that the proposed method was effective. Additionally, we confirmed that the degree of appropriateness was useful for response timing detection.

DOI： 10.14923/transfunj.2022hap0013

Identifying Influential References in Scholarly Papers Using Citation Contexts

Bottom-up Japanese Word Ordering using BERT Reviewed

Masato Yamazoe, Tomohiro Ohno and Shigeki Matsubara

Proceedings of the 15th International Conference on Agents and Artificial Intelligence (ICAART-2023) Vol. 3 page： 673 - 681 2023.2

Japanese Word Reordering based on Topological Sort Reviewed

Peng Sun, Tomohiro Ohno and Shigeki Matsubara

Proceedings of the 15th International Conference on Agents and Artificial Intelligence (ICAART-2023) Vol. 3 page： 768 - 775 2023.2

Question Difficulty Decision for Automated Interview in Foreign Language Test Reviewed

Minghao Lin, Koichiro Ito and Shigeki Matsubara

Proceedings of the 15th International Conference on Agents and Artificial Intelligence (ICAART-2023) Vol. 3 page： 768 - 775 2023.2

Ikoma T., Matsubara S.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol. 14458 page： 152 - 161 2023

Japanese Word Reordering Based on Topological Sort Open Access

Sun P., Ohno T., Matsubara S.

International Conference on Agents and Artificial Intelligence Vol. 3 page： 768 - 775 2023

Language：English Publishing type：Research paper (scientific journal) Publisher：International Conference on Agents and Artificial Intelligence

In Japanese, some sentences are grammatically well-formed, but not easy to read. This paper proposes a method for Japanese word reordering, which first adapts a Japanese BERT model to predict suitable pairwise orderings between two words in a sentence, and then converts the predicted results into a graph. The vertices in the graph represent the words in the sentence. The edges represent the pairwise orderings between two words. Finally, topological sort is applied to find the correct word ordering in a sentence by visiting each graph vertex. We conducted an evaluation experiment with uneasy-to-read Japanese sentences created from newspaper article sentences.

DOI： 10.5220/0011775000003393

Question Difficulty Decision for Automated Interview in Foreign Language Test Open Access

Lin M., Ito K., Matsubara S.

International Conference on Agents and Artificial Intelligence Vol. 3 page： 760 - 767 2023

Language：English Publishing type：Research paper (scientific journal) Publisher：International Conference on Agents and Artificial Intelligence

As globalization continues to progress all over the world, demand is growing for objective and rapid assessment of language proficiency in foreign language learners. While automated assessment of listening, reading, and writing skills has been proposed, little research has been done to automate assessment of speaking skills. In this paper, we propose a method of deciding the difficulty of questions generated for interview tests of skills assessment in speaking a foreign language. To address question difficulty flexibly according to the abilities of test takers, our method considers the appropriateness of responses from test takers. We implemented this method using the large-scale pre-trained language model BERT (Bidirectional Encoder Representation from Transformers). Experiments were conducted using simulated test data from the Japanese Learner’s Conversation Database to confirm the effectiveness of our method in deciding difficulty.

DOI： 10.5220/0011774900003393

Bottom-up Japanese Word Ordering Using BERT Open Access

Yamazoe M., Ohno T., Matsubara S.

International Conference on Agents and Artificial Intelligence Vol. 3 page： 673 - 681 2023

Language：English Publishing type：Research paper (scientific journal) Publisher：International Conference on Agents and Artificial Intelligence

Although Japanese is widely regarded as a relatively free word order language, word order in Japanese is not entirely arbitrary and has some sort of preference. As a result, it is an important technique to produce a sentence that is not only grammatically correct but also easy to read. This paper proposes a method for the word ordering of a whole Japanese sentence when the dependency relations between words are given. Using BERT, the method identifies the easy-to-read word order for a syntactic tree based on bottom-up processing. We confirmed the effectiveness of our method through an experiment on word ordering using newspaper articles.

DOI： 10.5220/0011743600003393

Construction of Responsive Utterance Corpus for Attentive Listening Response Production Reviewed

Extracting Information about Research Resources from Scholarly Papers Reviewed

Ayahito Saji and Shigeki Matsubara

Proceedings of the 24th International Conference on Asia-Pacific Digital Libraries (ICADL-2022) 2022.12

Classification of URL Citations in Scholarly Papers for Promoting Utilization of Research Artifacts Reviewed

Masaya Tsunokake and Shigeki Matsubara

Proceedings of AACL-IJCNLP-2022 The 1st Workshop on Information Extraction from Scientific Publications 2022.11

A Model-Theoretic Formalization of Natural Language Inference Using Neural Network and Tableau Method Reviewed

Ayahito Saji, Yoshihide Kato, Shigeki Matsubara

Proceedings of the 36th Pacific Asia Conference on Language, Information and Computation (PACLIC 35) 2022.10

Ito K., Murata M., Ohno T., Matsubara S.

2022 Language Resources and Evaluation Conference, LREC 2022 page： 7244 - 7252 2022

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings) Publisher：2022 Language Resources and Evaluation Conference, LREC 2022

In Japan, the number of single-person households, particularly among the elderly, is increasing. Consequently, opportunities for people to narrate are being reduced. To address this issue, conversational agents, e.g., communication robots and smart speakers, are expected to play the role of the listener. To realize these agents, this paper describes the collection of conversational responses by listeners that demonstrate attentive listening attitudes toward narrative speakers, and a method to annotate existing narrative speech with responsive utterances is proposed. To summarize, 148,962 responsive utterances by 11 listeners were collected in a narrative corpus comprising 13,234 utterance units. The collected responsive utterances were analyzed in terms of response frequency, diversity, coverage, and naturalness. These results demonstrated that diverse and natural responsive utterances were collected by the proposed method in an efficient and comprehensive manner. To demonstrate the practical use of the collected responsive utterances, an experiment was conducted, in which response generation timings were detected in narratives.

Extracting Information about Research Resources from Scholarly Papers

Saji, A; Matsubara, S

FROM BORN-PHYSICAL TO BORN-VIRTUAL: AUGMENTING INTELLIGENCE IN DIGITAL LIBRARIES, ICADL 2022 Vol. 13636 page： 440 - 448 2022

This paper presents a method to extract information about research resource metadata from scholarly papers to expand research resource repositories using an academic knowledge graph. Here, we considered the hypothesis that information extracted from scholarly papers is beneficial to expand research resource repositories from two perspectives: (1) the amount of information in existing metadata, and (2) the number of entries in existing repositories. We constructed an academic knowledge graph, where the nodes and directed edges in the graph correspond to the entities and their relations in scholarly papers, respectively. To verify our hypothesis, we constructed a knowledge graph using 15,721 papers published in international conferences. We then investigated the expandability of a language resource metadata repository using the constructed knowledge graph. The experimental results demonstrated that the constructed knowledge graph could be used to enrich the descriptions of metadata and increase the number of entries in the repository.

DOI： 10.1007/978-3-031-21756-2_35

Web of Science

Construction of Responsive Utterance Corpus for Attentive Listening Response Production

Ito, K; Murata, M; Ohno, T; Matsubara, S

LREC 2022: THIRTEEN INTERNATIONAL CONFERENCE ON LANGUAGE RESOURCES AND EVALUATION page： 7244 - 7252 2022

Estimating the Generation Timing of Responsive Utterances by Active Listeners of Spoken Narratives Reviewed

Koichiro Ito, Masaki Murata, Tomohiro Ohno, and Shigeki Matsubara

Proceeding of 2021 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU 2021) page： 495 - 502 2021.12

Using Citation Contexts in Scholarly Papers for Research Data Search Reviewed

Masaya Tsunokake and Shigeki Matsubara

Proceedings of the 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP 2021) 2021.12

Identification of Important Utterances in Narrative Speech Using Attentive Listening Responses Reviewed

Shunsuke Kosugi, Koichiro Ito, Masaki Murata, Tomohiro Ohno, and Shigeki Matsubara

Proceedings of the 16th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP 2021) 2021.12

A New Representation for Span-based CCG Parsing Reviewed

Yoshihide Kato and Shigeki Matsubara

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP-2021) page： 10579 - 10584 2021.11

Natural Language Inference using Neural Network and Tableau Method Reviewed

Ayahito Saji, Daiki Takao, Yoshihide Kato and Shigeki Matsubara

Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation (PACLIC 35) page： 406 - 414 2021.11

同時通訳における順送り方略の適用に影響する起点言語の特徴 Reviewed

蔡仲煕, 笠浩一朗, 松原茂樹

通訳翻訳研究 Vol. 20 page： 25 - 40 2021.2

Word reordering and comma insertion integrated with shift-reduce dependency parsing Reviewed

Miyachi K., Ohno T., Matsubara S.

ICAART 2021 - Proceedings of the 13th International Conference on Agents and Artificial Intelligence Vol. 2 page： 1144 - 1150 2021.2

Language：Japanese Publishing type：Research paper (international conference proceedings) Publisher：ICAART 2021 - Proceedings of the 13th International Conference on Agents and Artificial Intelligence

Japanese has widely recognized as relatively free word order language. However, since Japanese word order is not completely arbitrary and has some sort of preference, even native Japanese writers often write Japanese sentences that are grammatically well-formed but not easy to read. Furthermore, in Japanese sentences, a comma plays an important role in explicitly separating the constituents such as words and phrases. This paper proposes a method of word reordering and comma insertion for hard-to-read Japanese sentences so that they become easier to read, as basic technique. Our contribution is to show the feasibility of concurrently performing word reordering, comma insertion, and dependency parsing.

A New Representation for Span-based CCG Parsing

Kato, Y; Matsubara, S

2021 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP 2021) page： 10579 - 10584 2021

Classification of URLs Citing Research Artifacts in Scholarly Documents based on Distributed Representations International journal

Tsunokake M., Matsubara S.

CEUR Workshop Proceedings of 2nd Workshop on Extraction and Evaluation of Knowledge Entities from Scientific Documents (EEKE2021) collocated with ACM/IEEE Joint Conference on Digital Libraries (JCDL2021) Vol. 3004 page： 20 - 25 2021

ESTIMATING THE GENERATION TIMING OF RESPONSIVE UTTERANCES BY ACTIVE LISTENERS OF SPOKEN NARRATIVES

Ito, K; Murata, M; Ohno, T; Matsubara, S

2021 IEEE AUTOMATIC SPEECH RECOGNITION AND UNDERSTANDING WORKSHOP (ASRU) page： 495 - 502 2021

Language：English Publishing type：Research paper (scientific journal) Publisher：2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021 - Proceedings

Narrating is a fundamental human desire. In order to satisfy the desire, it is more effective to have multiple listeners listen to narratives than to have one listener does. Conversational agents, e.g., communication robots and smart speakers, are expected to listen to narratives, instead of human listeners. For such agents to be recognized as narrative listeners, it is necessary to generate responsive utterances at appropriate timings. This paper proposes a method to estimate the generation timings of responsive utterances by active listeners of spoken narratives. The proposed method estimates them by encoding acoustic and linguistic information extracted from narratives with a transformer-based approach. To evaluate the performance of the proposed method, an experiment was conducted using 131, 616 responsive utterances to narratives. The experimental results demonstrated that the proposed method was effective for the estimation and both the acoustic and linguistic information was useful.

DOI： 10.1109/ASRU51503.2021.9688132

Web of Science

Natural Language Inference using Neural Network and Tableau Method

Saji A., Takao D., Kato Y., Matsubara S.

Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021 page： 303 - 311 2021

Language：English Publishing type：Research paper (scientific journal) Publisher：Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021

Natural language inference (NLI) is the task of identifying the inferential relation between a text pair. Neural-based approaches recently achieved high performance on the NLI task, but they cannot explain their reasoning processes. On the other hand, symbolic approaches have an advantage that the reasoning process is understandable to humans. This paper proposes a method to integrate a neural NLI model and a tableau proof system, the latter of which explains the reasoning processes. The standard tableau method consists of decomposing logical formulas by applying inferential rules, and checking whether or not there exists a valuation that satisfies the given constraints. Unlike the standard tableau method, our method adopts not logical formulas but dependency structures as its ingredients and uses a neural NLI model for the latter process. In order to analyze the behavior of a neural NLI model, we conducted an experiment on the neural NLI model alone and the proposed method using the SNLI corpus.

A New Representation for Span-based CCG Parsing

Kato Y., Matsubara S.

EMNLP 2021 - 2021 Conference on Empirical Methods in Natural Language Processing, Proceedings page： 10579 - 10584 2021

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：EMNLP 2021 - 2021 Conference on Empirical Methods in Natural Language Processing, Proceedings

This paper proposes a new representation for CCG derivations. CCG derivations are represented as trees whose nodes are labeled with categories strictly restricted by CCG rule schemata. This characteristic is not suitable for span-based parsing models because they predict node labels independently. In other words, span-based models may generate invalid CCG derivations that violate the rule schemata. Our proposed representation decomposes CCG derivations into several independent pieces and prevents the span-based parsing models from violating the schemata. Our experimental result shows that an off-the-shelf span-based parser with our representation is comparable with previous CCG parsers.

Word Reordering and Comma Insertion Integrated with Shift-Reduce Dependency Parsing

Miyachi, K; Ohno, T; Matsubara, S

ICAART: PROCEEDINGS OF THE 13TH INTERNATIONAL CONFERENCE ON AGENTS AND ARTIFICIAL INTELLIGENCE - VOL 2 page： 1144 - 1150 2021

What Affects the Word Order of Target Language in Simultaneous Interpretation

Zhongxi Cai, Koichiro Ryu, and Shigeki Matsubara

2020 International Conference on Asian Language Processing, IALP 2020 page： 135 - 140 2020.12

Identification and Classification of Research Data Cited in Scholarly Papers

Tsunokake Masaya, Matsubara Shigeki

IEEJ Transactions on Electronics, Information and Systems Vol. 140 ( 12 ) page： 1357 - 1364 2020.12

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Institute of Electrical Engineers of Japan

This paper proposes a method for identifying and classifying the research data cited in scholarly papers, aiming at automatic generation of metadata stored in data repository. This study focuses on URL citations in the scholarly papers. That is, the targets are to identify the URLs referring to the research data and to classify them into tool and data. The method is realized as a multi-class classification (tool/data/others). The method acquires the distributed representations of the URLs from the context around them, and uses them as the input feature. There exists an advantage in that the meanings of URLs can be given based on their surrounding words. This study adopts an approach of computing the meaning of the entire URL from those of the components of the URL. In order to evaluate the performance of the proposed method, experiments on URL classification were conducted. The scholarly papers included in the proceedings of the international conference were used as experimental data. Experimental results have shown the effectiveness of the proposed method for identifying and classifying URLs referring to research data.

DOI： 10.1541/ieejeiss.140.1357

Identification of Research Data References Based on Citation Contexts

Parsing Gapping Constructions Based on Grammatical and Semantic Roles Reviewed

Yoshihide Kato, Shigeki Matsubara

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP-2020) page： 2747 - 2752 2020.11

Tomoki Ikoma, Shigeki Matsubara

Vol. 12504 LNCS page： 149 - 156 2020

Relation between degree of empathy for narrative speech and type of responsive utterance in attentive listening

Koichiro Ito, Masaki Murata, Tomohiro Ohno, Shigeki Matsubara

LREC 2020 - 12th International Conference on Language Resources and Evaluation, Conference Proceedings page： 696 - 701 2020

Features of Source Language That Affect the Word Order of Target Language in Simultaneous Interpretation Reviewed

CAI Zhongxi, RYU Koichiro, MATSUBARA Shigeki

Interpreting and Translation Studies: The Journal of the Japan Association for Interpreting and Translation Studies Vol. 20 ( 0 ) page： 25 - 40 2020

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Japan Association for Interpreting and Translation Studies

Simultaneous interpretation has to be conducted to shorten delay and reduce working memory
load. Interpreters adopt a strategy that involves the generation of a target language in the word
order that is similar to that of the source language. To clarify how the interpreters decide the
word order in practice, we conduct a statistical study on the word order of simultaneous
interpretation, based on comparison with written translation. The word order comparisons
indicate that the strategy of maintaining the original word order is more likely to be adopted in
cases where the word order is not consistent in different translations. We analyze the features of
source languages, including dependencies of chunk pairs, length of post modifier, and distance
between two chunks, and confirm that such features affect the word-order decision in
simultaneous interpretation.

DOI： 10.50837/its.2002

PTB graph parsing with tree approximation Reviewed International journal

Kato Y., Matsubara S.

ACL 2019 - 57th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference page： 5344 - 5349 2020

Authorship：Last author Language：English Publishing type：Research paper (international conference proceedings) Publisher：ACL 2019 - 57th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference

The Penn Treebank (PTB) represents syntactic structures as graphs due to nonlocal dependencies. This paper proposes a method that approximates PTB graph-structured representations by trees. By our approximation method, we can reduce nonlocal dependency identification and constituency parsing into single tree-based parsing. An experimental result demonstrates that our approximation method with an off-the-shelf tree-based constituency parser significantly outperforms the previous methods in nonlocal dependency identification.

Relation between Degree of Empathy for Narrative Speech and Type of Responsive Utterance in Attentive Listening

What Affects the Word Order of Target Language in Simultaneous Interpretation

Cai Zhongxi, Ryu Koichiro, Matsubara Shigeki

2020 INTERNATIONAL CONFERENCE ON ASIAN LANGUAGE PROCESSING (IALP 2020) page： 135 - 140 2020

Ito, K; Murata, M; Ohno, T; Matsubara, S

PROCEEDINGS OF THE 12TH INTERNATIONAL CONFERENCE ON LANGUAGE RESOURCES AND EVALUATION (LREC 2020) page： 696 - 701 2020

Parsing Gapping Constructions Based on Grammatical and Semantic Roles

Kato, Y; Matsubara, S

PROCEEDINGS OF THE 2020 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP) page： 2747 - 2752 2020

Parsing gapping constructions based on grammatical and semantic roles

Kato Y., Matsubara S.

EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference page： 2747 - 2752 2020

Language：English Publishing type：Research paper (scientific journal) Publisher：EMNLP 2020 - 2020 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference

A gapping construction consists of a coordinated structure where redundant elements are elided from all but one conjuncts. This paper proposes a method of parsing sentences with gapping to recover elided elements. The proposed method is based on constituent trees annotated with grammatical and semantic roles that are useful for identifying elided elements. Our method outperforms the previous method in terms of F-measure and recall.

Statistical analyses of missing translations in simultaneous interpretation using a large-scale bilingual speech corpus

PTB Graph Parsing with Tree Approximation Reviewed

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL-2019) page： 5344-5349 2019.7

同時通訳における作業記憶の負荷と訳出との関係 Reviewed

笠浩一朗, 松原茂樹

通訳翻訳研究 ( 18 ) page： 1-12 2019.2

同時通訳における語の欠落に影響を及ぼす要因の分析 Reviewed

蔡仲煕, 笠浩一朗, 松原茂樹

通訳翻訳研究 ( 18 ) page： 1-14 2019.2

Cai Z., Ryu K., Matsubara S.

LREC 2018 - 11th International Conference on Language Resources and Evaluation page： 4282 - 4288 2019

Language：English Publishing type：Research paper (scientific journal) Publisher：LREC 2018 - 11th International Conference on Language Resources and Evaluation

This paper describes statistical analyses of missing translations in simultaneous interpretations. Eighty-eight lectures from English-to-Japanese interpretation data from a large-scale bilingual speech corpus were used for the analyses. Word-level alignment was provided manually, and English words without corresponding Japanese words were considered missing translations. The English lectures contained 46,568 content words, 33.1% of which were missing in the translation. We analyzed the relationship between missing translations and various factors, including the speech rate of the source language, delay of interpretation, part-of-speech, and depth in the syntactic structure of the source language. The analyses revealed that the proportion of missing translations is high when the speech rate is high and delay is large. We also found that a high proportion of adverbs were missed in the translations, and that words at deeper positions in the syntactic structure were more likely to be missed.

PTB Graph Parsing with Tree Approximation

Kato, Y; Matsubara, S

57TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2019) page： 5344 - 5349 2019

Model-Theoretic Incremental Interpretation Based on Discourse Representation Theory Reviewed

Yoshihide Kato, Shigeki Matsubara

Proceedings of the 32nd Pacific Asia Conference on Language, Information and Computation (PACLIC 32) page： 1 2018.12

Statistical Analysis of Missing Translation in Simultaneous Interpretation Using A Large-scale Bilingual Speech Corpus Reviewed

Zhongxi Cai, Koichiro Ryu, Shigeki Matsubara

Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018) page： 4282-4288 2018.5

Collection of Responsive Utterances to Show Attentive Hearing of Narratives

Murata Masaki, Ohno Tomohiro, Matsubara Shigeki

IEEJ Transactions on Electronics, Information and Systems Vol. 138 ( 5 ) page： 637 - 638 2018.5

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Institute of Electrical Engineers of Japan

This paper describes collection of responsive utterances of showing attentive hearing for narrative speech. In this research, a method of annotating an existing narrative corpus with responsive utterances was adopted. Actually, 16,213 responsive utterances were collected for a narrative corpus consisting of 13,381 utterances. It was confirmed that the method can collect a variety of natural responsive utterances.

DOI： 10.1541/ieejeiss.138.637

Corpus-based Analysis of Factors Leading to Missing Translation in Simultaneous Interpretation

CAI Zhongxi, RYU Koichiro, MATSUBARA Shigeki

Interpreting and Translation Studies: The Journal of the Japan Association for Interpreting and Translation Studies Vol. 18 ( 0 ) page： 133 - 146 2018

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Japan Association for Interpreting and Translation Studies

This paper describes corpus-based analyses of missing translations in simultaneous interpretations. We used English-to-Japanese interpretation data in the simultaneous interpretation database of Nagoya University for the analyses. The English lectures contained 46,568 content words, and 33.1% of them were missing in the translation. We analyzed the relations between omissions and the factors that influence them, i.e., the speech rate of the source language, delay of interpretation, part-of-speech, and depth in the syntactic structure of the source language. The analyses revealed that the proportion of missing translations is high when the speech rate is high, and the delay of interpretation is large, i.e., adverbs were missing in the translations in a high proportion; the words at deeper positions in the syntactic structure were more likely to be missed.

DOI： 10.50837/its.1808

Effect of Translating on Working Memory Load in Simultaneous Interpreting International journal

RYU Koichiro, MATSUBARA Shigeki

Interpreting and Translation Studies: The Journal of the Japan Association for Interpreting and Translation Studies Vol. 18 ( 0 ) page： 147 - 158 2018

Language：Japanese Publishing type：Research paper (scientific journal) Publisher：The Japan Association for Interpreting and Translation Studies

By employing corpus-based analyses of English-Japanese simultaneous interpretation, we discuss the influence translations have on the working memory load of simultaneous interpreters. To conduct a quantitative analysis with a large corpus, we used the Simultaneous Interpretation Database of Nagoya University. We analyzed a total of 88 sets of simultaneous interpretations of 22 different lectures in English given by 17 interpreters. When the load on the working memory of simultaneous interpreters is increased, the speech time increases, the translation rate decreases, but the average speech rate does not exhibit a relevant change.

DOI： 10.50837/its.1809

Model-theoretic incremental interpretation based on discourse representation theory

Kato Y., Matsubara S.

Proceedings of the 32nd Pacific Asia Conference on Language, Information and Computation, PACLIC 2018 page： 254 - 262 2018

Language：English Publishing type：Research paper (scientific journal) Publisher：Proceedings of the 32nd Pacific Asia Conference on Language, Information and Computation, PACLIC 2018

This paper proposes a model-theoretic approach to incremental interpretation where allsentence prefixes have semantic values. Theproposed semantics is based on DiscourseRepresentation Theory (DRT), where semantic representations (called DRSs) are interpreted as assignment updates. In our semantics, a partial DRS of a sentence prefix is interpreted as two sets which stipulate the assignment updates. One denotes possible updatesand the other denotes necessary updates. Withthe proposed semantics, we can assign truthvalues to sentence prefixes.

Statistical Analyses of Missing Translations in Simultaneous Interpretation Using a Large-scale Bilingual Speech Corpus

Cai, ZX; Ryu, K; Matsubara, S

PROCEEDINGS OF THE ELEVENTH INTERNATIONAL CONFERENCE ON LANGUAGE RESOURCES AND EVALUATION (LREC 2018) page： 4282 - 4288 2018

Correcting Syntactic Annotation Errors Based on Tree Mining Reviewed

Kanta Suzuki, Yoshihide Kato, Shigeki Matsubara

IEICE Transactions on Information and Systems Vol. E100-D ( 5 ) page： 1106-1113 2017.5

Collection of responsive utterances to show attentive hearing attitude to speakers

Tomohiro Ohno, Masaki Murata, Shigeki Matsubara

Proceedings of the 11th International Conference on Ubiquitous Information Management and Communication (IMCOM-2017) ( 31 ) 2017.1

対話コーパスを用いた相づち生成タイミングの検出 Reviewed

大野誠寛，神谷優貴，松原茂樹

電子情報通信学会論文誌 Vol. J100-A ( 1 ) page： 53-65 2017.1

Left-Corner Parsing for Identifying PTB-Style Nonlocal Dependencies

Kato Yoshihide, Matsubara Shigeki

Journal of Natural Language Processing Vol. 24 ( 3 ) page： 371 - 394 2017

Language：English Publishing type：Research paper (scientific journal) Publisher：The Association for Natural Language Processing

Nonlocal dependencies represent syntactic phenomenon such as wh-movement, A-movement in passives, topicalization, raising, control, and right node raising. Nonlocal dependencies play an important role in semantic interpretation. This paper proposes a left-corner parser that identifies nonlocal dependencies. Our parser integrates nonlocal dependency identification into a transition-based system. We adopt a left-corner strategy in order to use the syntactic relation c-command, which plays an important role in nonlocal dependency identification. To utilize the global features captured by nonlocal dependencies, our parser uses a structured perceptron. In experimental evaluations, our parser achieved a good balance between constituent parsing and nonlocal dependency identification.

DOI： 10.5715/jnlp.24.371