# 茂木光志 — 論文
## 学内文書群を活用した生成AIによる研究データ管理ガイドラインの生成
茂木光志, 松原茂樹
大学ICT推進協議会 2026年度 年次大会 \(AXIES2026\) (2026)
学術機関の研究データガバナンスを構築するために,研究データ管理ガイドラインを策定することは重要である.しかし,大学における研究データ管理ガイドラインは,規程やポリシー,サービスマニュアルなど既存の学内文書と整合させる必要があり,その策定には膨大な労力を要する.本論文では,研究データに関わる既存の学内文書群を用いた研究データ管理ガイドラインの生成方式について述べる.本方式では,策定するガイドラインの構成に基づき,生成AIを用いてガイドラインの本文を生成する.生成の過程における学内文書の利用形式の違いにより複数の方式を試行する.名古屋大学の学内文書群,ならびに,策定された研究データ管理ガイドラインの章構成を用いて,ガイドライン本文の生成実験を行った.
## 生成AIを用いた研究データのメタデータ生成とリポジトリへの登録:論文におけるURL引用の分類とREADMEの活用
岡田怜真, 茂木光志, 松原茂樹
大学ICT推進協議会 2026年度 年次大会 \(AXIES2026\) (2026)
本論文では,研究データのメタデータ自動生成の新たな方式を提案する.機関リポジトリを介した研究データ公開は,共通メタデータを作成する負担もあり,必ずしも加速化されているとはいえない.本方式では,研究データと紐づく学術論文,ならびに,研究データの説明文書であるREADMEを情報源に,生成AIを用いてメタデータを生成する.また,論文テキストに出現するURL引用を自動分類することで,学術論文と紐づく研究データを同定する.データプラットフォームで公開されている研究データを対象に実施した実験では,論文テキストから59,650件,READMEから41,539件のメタデータが生成され,大規模データリポジトリの効率的な生成可能性が示された.
## What, Why, and How Research Data Are Cited? Investigating Research Data Citation Practices Across Time and Domains
Koshi Motegi, Koichiro Ito, Shigeki Matsubara
The 28th International Conference on Asia-Pacific Digital Libraries \(ICADL 2026\) (2026)
## Where Are Research Data Published? Identifying Their Publication URLs in Scholarly Papers
Ryoma Okada, Koshi Motegi, Koichiro Ito, Shigeki Matsubara
The 28th International Conference on Asia-Pacific Digital Libraries \(ICADL 2026\) (2026)
## Cross-Format Citation Classification: Examining the Effects of Combining Bibliographic and URL Citations
Koshi Motegi, Koichiro Ito, Shigeki Matsubara
Proceedings of the 26th ACM/IEEE Joint Conference on Digital Libraries \(JCDL 2026\), pp. 1–6, 2026.10 (2026)
[DOI: 10.1145/3805696.3846499](https://doi.org/10.1145/3805696.3846499)
[掲載元](https://dl.acm.org/doi/10.1145/3805696.3846499)
Citations in scholarly papers serve various functions and refer to diverse resources. Methods have been proposed to classify citation functions and the types of cited resources. In scholarly papers, citations take two formats, i.e., citations using citation tags \(bibliographic citations\) and citations using URLs \(URL citations\). To utilize information about citation functions and the types of cited resources, it is important to be able to classify them across both formats; however, previous studies have classified them independently for each format. This study investigated the effect of combining bibliographic and URL citations on the classification of citation functions and the types of cited resources. We constructed classifiers through transfer learning and multitask learning, and evaluated their performance. The results demonstrated that combining the two citation formats improved classification performance across various encoder models and learning approaches. The improvements were particularly large for cited resource type classification of bibliographic citations.
## Identifying Implicit Research Data References in Paper Citations
Koshi Motegi, Koichiro Ito, Shigeki Matsubara
Proceedings of the 3rd International Workshop on Natural Scientific Language Processing \(NSLP 2026\) co-located with LREC 2026, pp. 186–192, 2026.05 (2026)
[DOI: 10.63317/2g9fq97f2h2j](https://doi.org/10.63317/2g9fq97f2h2j)
[掲載元](https://aclanthology.org/2026.nslp-1.18/)
To encourage the public release of research data under open science, it is beneficial to establish mechanisms for evaluating research data based on metrics such as citation counts. In scholarly papers, authors sometimes cite papers that report the creation or release of research data instead of citing the research data themselves. In this paper, as a step toward computing citation counts of research data, we investigate the feasibility of identifying paper citations that refer to research data. We conducted an identification experiment using large language models and evaluated their performance.
## Where Did the Research Data Originate? Acquiring Provenance of Research Data from Scholarly Papers
Koshi Motegi, Koichiro Ito, Shigeki Matsubara
The 27th International Conference on Asia-Pacific Digital Libraries \(ICADL 2025\) (2025)
[DOI: 10.1007/978-981-95-4861-3\_29](https://doi.org/10.1007/978-981-95-4861-3_29)
[掲載元](https://link.springer.com/chapter/10.1007/978-981-95-4861-3_29)
Prior knowledge of existing methodologies and resources is essential for a comprehensive understanding of research. In scholarly papers, citations help readers identify prior knowledge. However, research data, such as datasets, lack such mechanisms, making it difficult to trace their origins. In this study, we discuss approaches to acquiring information about the origins of research data, i.e., research data provenance. We focused on scholarly papers as a source for acquiring information about research data provenance, as they often include descriptions of how research data were created. Accordingly, we conducted a preliminary experiment on extracting information about research data provenance from scholarly papers using large language models \(LLMs\). The results demonstrated the feasibility of extracting information about research data provenance from scholarly papers.
## 語り手の発話の言い換えにより語りに傾聴を示す応答の生成
茂木光志, 伊藤滉一朗, 村田匡輝, 松原茂樹
第31回言語処理学会年次大会 \(NLP2025\) (2025)
[掲載元](https://www.anlp.jp/proceedings/annual_meeting/2025/pdf_dir/P10-10.pdf)
語る機会を創出するために,会話エージェントが語りの聴き手を担うことが期待されている.これらが聴き手として認められるためには,語り手に対して傾聴態度を示す発話(傾聴応答)を生成することが有効である.傾聴応答の 1 つに,語り手の発話を言い換える応答(言い換え応答)がある.言い換え応答を適切に生成できれば,語りを理解していることを示すことに寄与する.本論文では,大規模言語モデルによる言い換え応答の生成可能性を検証する.生成実験を行い,その性能を評価した.
## 研究データのオープンアクセスを加速化する:生成AIを用いたメタデータ生成と機関リポジトリへの収載
渡邉優, 茂木光志, 松原茂樹
大学ICT推進協議会 2025年度 年次大会 \(AXIES2025\) (2025)
[DOI: 10.24669/axies.2025.0\_702](https://doi.org/10.24669/axies.2025.0_702)
[掲載元](https://www.jstage.jst.go.jp/article/axies/2025/0/2025_702/_article/-char/ja)
本論文では,大学における研究データのオープンアクセスの加速化を目指し,研究データのメタデータ生成とその機関リポジトリへの収載の新たな方式を提案する.大学における研究データ公開は,メタデータ作成の負担が小さくなく,十分に進んでいないのが現状である.本方式では,学術論文テキストおよび公開プラットフォーム内の情報を活用し,大規模言語モデル(LLM)によりメタデータを自動生成し,それをリポジトリソフトウェアに搭載する.論文テキストデータを用いた実験では,同定された16,426件の研究データの64.8%に相当する10,640件についてメタデータが生成され,データリポジトリの効率的作成の実現性が示された.
大学ICT推進協議会2025年度年次大会(AXIES2025)最優秀論文賞
## 学術論文において研究データを参照する論文引用の識別
茂木 光志, 伊藤 滉一朗, 松原 茂樹
言語処理学会第32回年次大会 \(NLP2026\) (2026)
[掲載元](https://www.anlp.jp/proceedings/annual_meeting/2026/pdf_dir/Q9-21.pdf)
オープンサイエンスの一環である研究データの公開促進のために,被引用数などに基づき研究データを評価できる仕組みを整えることが考えられる.学術論文では,研究データの引用の代替として,その作成や公開を報じる論文が引用されることがある.本論文では,研究データの被引用数の算出に向けて,研究データを参照する論文引用を識別することの実現性について考察する.LLM を用いた識別実験を行い,その性能を評価した.
## 研究活動で産出された論文と研究データの対応付け
岡田 怜真, 茂木 光志, 伊藤 滉一朗, 松原 茂樹
言語処理学会第32回年次大会 \(NLP2026\) (2026)
[掲載元](https://www.anlp.jp/proceedings/annual_meeting/2026/pdf_dir/Q9-22.pdf)
オープンサイエンスの潮流のもと,論文と研究データの公開が進んでいる.同一の研究活動で産出された論文と研究データが相互に参照可能であることは,双方の理解を深めるうえで重要となる.本論文では,同一の研究活動で産出された論文と研究データを対応付ける手法について述べる.本手法では,論文中の URL を,その周辺テキストを利用して,対応する研究データの公開先であるか否かに分類する.BERT ベースのモデルを fine-tuning することで本手法を実装し,実験でその性能を評価した.