(日本語はこちら)
Hi! My name is Tatsuya Ishigaki (石垣達也), currently a researcher on Natural Language Processing (NLP) at Artificial Intelligence Research Center, AIST, Japan.
My research topics are about NLP technologies to lower language barriers. Recent main interests are neural netowork-based approaches for tasks such as:
Automatic Commentary Generation [INLG2021, EMNLP2022, [INLG2023 (best demo paper award)], [NL250 (best paper award)]
Automatic summarization [ECIR2020], [ECIR2020], [LREC2022], [NL240 (best paper award)][NLP2019 (young researcher award)]
Appying NLP to read-world problems, e.g., scenario planning and future risk analysis. [LREC2022]
Low-resource NLP (mainly targeting Japanese) [RANLP2023],[PACLIC2023]
Please feel free to email me if you are interested in collaboration with me!
Contact: ishigaki.tatsuya@aist.go.jp, Google Scholar, LinkedIn, GitHub, Twitter,
Work Experiences
AIRC (November, 2020 - present) : as a researcher.
AIRC (April, 2020 - November 2020, October) : as a postdoc.
Okumura laboratory at Tokyo Institute of Technology(April, 2019 - March, 2020): as a researcher.
AIRC (April, 2018 - March, 2019) : as a technical staff.
Okumura laboratory (April, 2018 - March, 2019): as a research assistant.
Ibaraki National Institute of Technology (April, 2015 - March, 2016): as a part-time lecturer.
Education
Ph..D: April, 2015 - September, 2019 Tokyo Institute of Technology, supervised by Prof. Hiroya Takamura.
M.D.: September, 2018 - December, 2018: National Taiwan University (visiting research student) supervised by Prof. Hsin-Hsi Chen.
April, 2013 - March, 2015: Tokyo Institute of Technology M.S supervised by Prof. Hiroya Takamura.
Publications
International Conferences (Refereed)
NLP is a fast-growing field, thus, NLP researchers often focus more on publishing papers at international conferences rather than journals.
Prompting for Numerical Sequences: A Case Study in Market Comment Generation, Masayuki Kawarada, Tatsuya Ishigaki, Hiroya Takamura, LREC-COLING2024 (to appear)
Training Generative Question-Answering on Synthetic Data Obtained from an Instruct-tuned Model, Kosuke Takahashi, Takahiro Omi, Kosuke Arima and Tatsuya Ishigaki, PACLIC2023
Constructing a Japanese Business Email Corpus Based on Social Situations, Muxuan Liu, Tatsuya Ishigaki, Yusuke Miyao, Hiroya Takamura and Ichiro Kobayashi, PACLIC2023
Audio Commentary System for Real-Time Racing Game Play, Tatsuya Ishigaki, Goran Topić, Yumi Hamazono, Ichiro Kobayashi, Yusuke Miyao and Hiroya Takamura, INLG2023 (demo paper, best paper award)
Pretraining Language- and Domain-Specific BERT on Automatically Translated Text, Tatsuya Ishigaki, Yui Uehara, Goran Topic, Hiroya Takamura, RANLP2023
Validation of a Foresight Support System to Imagine an Uncertain Future:-Effectiveness Testing through Scenario Planning Workshops, Suzuko Nishino, Yuichi Washida, Tatsuya Ishigaki, Sohei Washino, Hiroki IGARASHI, Akihiko Murai, Yukari Nagai, KICSS2023
Open-Domain Video Commentary Generation, Edison Marrese-Taylor, Yumi Hamazono, Tatsuya Ishigaki, Goran Topic, Yusuke Miyao, Ichiro Kobayashi, Hiroya Takamura, EMNLP2022 [paper]
Automating Horizon Scanning in Future Studies, Tatsuya Ishigaki, Suzuko Nishino, Sohei Washino, Hiroki Igarashi, Yukari Nagai, Yuichi Washida and Akihiko Murai, LREC2022 [paper]
Unpredictable Attributes in Data-to-text Generation, Yumi Hamazono, Tatsuya Ishigaki, Ichiro Kobayashi, Yusuke Miyao and Hiroya Takamura, PACLIC2021 (acceptance rate (oral): 48%=52/110)
Generating Racing Game Commentary from Vision, Language, and Structured Data, Tatsuya Ishigaki, Goran Topic, Yumi Hamazono, Hiroshi Noji, Ichiro Kobayashi, Yusuke Miyao and Hiroya Takamura, INLG2021 [paper] (acceptance rate: 40% = 31/76)
Learning with Contrastive Examples for Data-to-Text Generation, Yui Uehara*, Tatsuya Ishigaki* (*equal contribution), Kasumi Aoki, Hiroshi Noji, Keiichi Goshima, Ichiro Kobayashi, Hiroya Takamura and Yusuke Miyao, COLING2020 (acceptance rate: 32.9% = 644/1956)
Neural Query-biased Abstractive Summarization Using Copying Mechanism, Tatsuya Ishigaki, Hen-Hsen Huang, Hiroya Takamura, Hsin-Hsi Chen, Manabu Okumura, ECIR2020 [paper][video] (acceptance rate: 30% = 46/159)
Distant Supervision for Extractive Question Summarization, Tatsuya Ishigaki, Kazuya Machida, Hayato Kobayashi, Hiroya Takamura, Manabu Okumura, ECIR2020 [paper][video] (acceptance rate: 30% = 46/159)
Semi-Supervised Extractive Question Summarization Using Question-Answer Pairs, *Kazuya Machida, *Tatsuya Ishigaki (*equal contribution), Hayato Kobayashi, Hiroya Takamura, Manabu Okumura, ECIR2020 [paper][video]. (acceptance rate: 30% = 46/159)
Controlling Contents in Data-to-Document Generation with Human-Designed Topic Labels, Kasumi Aoki, Akira Miyazawa, Tatsuya Ishigaki, Tatsuya Aoki, Hiroshi Noji, Keiichi Goshima, Ichiro Kobayashi, Hiroya Takamura and Yusuke Miyao, INLG2019 [paper] (acceptance rate: 24.5% = 36/147)
Discourse-Aware Hierarchical Attention Network for Extractive Single-Document Summarization, Tatsuya Ishigaki, Hidetaka Kamigaito, Hiroya Takamura and Manabu Okumura, RANLP2019 [paper] (acceptance rate: 26.7% = 37/134)
Learning to Select, Track, and Generate for Data-to-Text, Hayate Iso, Yui Uehara, Tatsuya Ishigaki, Hiroshi Noji, Eiji Aramaki, Ichiro Kobayashi, Yusuke Miyao, Naoaki Okazaki, Hiroya Takamura, ACL2019 [paper] (acceptance rate 25.7% = 447/1740)
Generating Market Comments Referring to External Resources, Tatsuya Aoki, Akira Miyazawa, Kasumi Aoki, Keiichi Goshima, Tatsuya Ishigaki, Ichiro Kobayashi, Hiroya Takamura and Yusuke Miyao, INLG2018 [paper] (acceptance rate: 60% = 62/102)
Summarizing Lengthy Questions, Tatsuya Ishigaki, Hiroya Takamura, and Manabu Okumura, IJCNLP2017 [paper] (acceptance rate: 31% = 103/337)
Journal
I usually write journal papers in Japanese for fomestinese readers by extending international conference papers. Please ask me if you are interested in the works below, then, I would be happy to share the original conference papers written in English.
ホライゾン・スキャニングの自動化のための言語処理応用, *西野 涼子,*石垣 達也,鷲野 壮平,五十嵐 広希,村井 昭彦,鷲田 祐一,永井 由佳里, 自然言語処理 Vol.30 No.3, 2023 (* equal contribution)
Controlling Contents in Data-to-Document Generation with Human-Designed Topic Labels, Kasumi Aoki, Akira Miyazawa, Tatsuya Ishigaki, Tatsuya Aoki, Hiroshi Noji, Keiichi Goshima, Ichiro Kobayashi, Hiroya Takamura and Yusuke Miyao, Computer Speech and Language Vol.66, 2020 [paper]
質問−回答ペアを活用する半教師あり抽出型質問要約モデルとその学習法 , 石垣 達也,町田 和哉, 小林 隼人, 高村 大也,奥村 学, 自然言語処理 Vol.27 No.4 (2020年12月号), 2020
Learning to Select, Track, and Generate for Data-to-Text, Hayate Iso, Yui Uehara, Tatsuya Ishigaki, Hiroshi Noji, Eiji Aramaki, Ichiro Kobayashi, Yusuke Miyao, Naoaki Okazaki and Hiroya Takamura, 自然言語処理 Vol.27 No.3, 2020 [paper]
複数文質問を対象とした抽出型および生成型要約, 石垣 達也,高村 大也,奥村 学, 自然言語処理 Volume 26 No.1 (2019年3月号), 2019 [paper]
Workshops / Domestic Conferences (written for Japanese readers)
I also write papers for domestic conferences in Japan to better discuss our studies with domestic academias and people from industry. After the discussions, most papers are extended to submitted versions for international conferences.
レーシングゲーム実況テキストモデリングのための運動力学的素性,石垣達也,上田 佳祐,トピチゴラン,小林一郎,宮尾祐介,高村大也,第253回自然言語処理研究会(NL253), 2022
一般ドメイン動画実況生成, 濵園侑美, 石垣達也, 宮尾祐介, 小林一郎, 高村大也, 第 36 回人工知能学会全国大会論文集 (JSAI2022), 2022
機械翻訳テキストを用いた低資源言語特定分野向けBERTの事前学習, 石垣達也, 上原由衣, トピチゴラン, 高村大也, 第21回情報科学技術フォーラム(FIT2022)
定義文自動生成による専門分野向けエンティティリンキングの精度向上, 石垣達也, 上原由衣, 劉珊珊, 松本裕治, 高村大也, 言語処理学会第28回年次大会 (NLP2022), 2022
実況発話ラベル予測モデルにおける状況認識素性の活用, 上田 佳祐, 石垣 達也, 小林 一郎, 宮尾 祐介, 高村 大也, 言語処理学会第28回年次大会(NLP2022), 2022
実況における発話ラベル予測, 上田 佳祐, 石垣 達也, 小林 一郎, 宮尾 祐介, 高村 大也, 第251回自然言語処理研究会 (NL251), 2021
レーシングゲーム実況生成, 石垣達也, トピチゴラン, 濵園 侑美, 能地 宏, 小林 一郎, 宮尾 祐介, 高村 大也, 第250回自然言語処理研究会 (NL250) (優秀研究賞), 2021
高分子材料に関する技術文献からの機械学習を用いた知識獲得, 石垣達也,上原 由衣,Liu Shanshan,Topic Goran,高村 大也, 第70回高分子討論会, 2021
解説: Learning with Contrastive Examples for data-to-text Generation, 上原由衣, 石垣達也, 自然言語処理 Volume 28 No.2 (解説記事), 2021
疑似負例を用いたdata-to-textモデルの学習, 上原 由衣, 石垣 達也 (equal contribution), 青木 花純, 能地 宏, 五島 圭一, 小林 一郎, 宮尾 祐介, 高村 大也, 第246回自然言語処理研究会 (NL246), 2020 [paper]
Supporting Information Recall for Elderly People in Hyper Aged Societies, Tatsuya Ishigaki, Jingyi You, Hiroki Takimoto, Manabu Okumura, HCII2020, 2020 [paper]
超高齢社会における高齢者のための情報想起支援システム, 石垣達也, You Jingyi, 瀧本 洋喜, 奥村 学, 言語処理学会第26回年次大会, 2020 [paper]
コピー機構を用いたクエリ指向ニューラル生成型要約, 石垣達也, Hen-Hsen Huang, Hsin-Hsi Chen, 高村 大也, 奥村 学, 第240回自然言語処理研究会(NL240)(優秀研究賞), 2019 [paper]
談話構造を考慮する階層的注意機構による抽出型ニューラル単一文書要約, 石垣 達也, 上垣外 英剛, 高村 大也, 奥村 学, 言語処理学会第25回年次大会(若手奨励賞), 2019 [paper][slide]
Data-to-Textにおける主題遷移のモデル化, 磯 颯, 上原 由衣, 石垣 達也, 能地 宏, 荒牧 英治, 小林 一郎, 宮尾 祐介, 岡崎 直観, 高村 大也, 言語処理学会第25回年次大会 , 2019 [paper]
Distant Supervisionによる質問要約, 石垣 達也, 町田 和哉, 小林 隼人, 高村 大也, 奥村 学, 第236回情報処理学会自然言語処理研究会(NL236), 2018
「長文質問」のための抽出型及び生成型要約, 石垣 達也, 高村 大也, 奥村 学, 第232回情報処理学会自然言語処理研究会(NL232), 2017
Software and Dataset Releases
材料科学分野BERT
Others
Awards
Best paper award (demo): INLG2023, 2023
優秀研究賞: 情報処理学会第250回自然言語処理研究会(NL250), 2021
優秀研究賞: 情報処理学会第240回自然言語処理研究会(NL240), 2019
若手奨励賞: 言語処理学会第25回年次大会(NLP2019), 2019
MEXT Leading Program "Academy of Global Leadership" (full tuition waiver(M.S and Ph.D) + 200,000JPY/month, October 2013 - March 2018)
Activities
Steering Commitee: Information Processing Society of Japan SIG-NL (2020-2024)
Local Chair: INLG2024
Program Committee: NLP2023, NLP2024
Reviewer: NAACL (2024), LREC-COLING (2024), EACL (2023), ACL (2023), EMNLP (2021, 2022,2023, 2024), ACL-IJCNLP (2021), AACL-IJCNLP (2022), Journal of Natural Language Processing (2020, 2021, 2022, 2023, 2024 (Editor))...