I am a senior researcher at Artificial Intelligence Research Center (AIRC), National Institute of Advanced Industrial Science and Technology (AIST), Japan, where I work on knowledge graph and semantic technology. I am also a collaborative associate professor at The University of Electro-Communications (UEC) ( Ohsuga, Tahara, & Sei Lab). I am also a part-time lecturer at Hosei University, where I teach semantic web (spring semester) and agent technology (fall semester).
I did my PhD at UEC, where I was advised by Takahiro Kawamura and Akihiko Ohsuga and funded by the JSPS DC2.
I'm interested in semantic web, ontology, graph representation learning, and open data. Much of my research is about constructing and reasoning knowledge graphs from physical and cyber worlds (unstructured text, semistructured data, video, virtual space, etc.).
|
18.
Fumika Okuhara, Shusaku Egami, Yuichi Sei, Yasuyuki Tahara, Akihiko Ohsuga Transactions of the Japanese Society for Artificial Intelligence, Vol.40, No.1, to appear, 2025 (in Japanese) to appear |
|
17.
Yuki Saito, Shusaku Egami, Yuichi Sei, Yasuyuki Tahara, Akihiko Ohsuga Transactions of the Japanese Society for Artificial Intelligence, Vol.39, No.6, pp.AG24-D_1-13, 2024 (in Japanese) DOI: https://doi.org/10.1527/tjsai.39-6_AG24-D This study investigated effective knowledge representation for anime recommendations using graph neural networks, demonstrating that combining metadata-based and text-based knowledge graphs significantly improves recommendation performance. |
|
16.
Yuki Sawamura, Takeshi Morita, Shusaku Egami, Takanori Ugai, Ken Fukuda Transactions of the Japanese Society for Artificial Intelligence, Vol.39, No.6, pp.C-O42_1-14, 2024 (in Japanese) DOI: hhttps://doi.org/10.1527/tjsai.39-6_C-O42 This study proposed a Japanese Entity Linking model using a Pointer Network and confirmed it outperformed existing multilingual models in evaluation experiments. |
|
15.
Swe Nwe Nwe Htun, Shusaku Egami, Ken Fukuda SICE Journal of Control, Measurement, and System Integration, Vol.17, No.1, pp.87-105, 2024 DOI: https://doi.org/10.1080/18824889.2024.2318848 This study investigates the potential of generating synthetic training data for activities of daily living (ADLs) recognition using the VirtualHome2KG framework. |
|
14.
Shusaku Egami, Takanori Ugai, Masateru Oota, Kyoumoto Matsushita, Takahiro Kawamura, Kouji Kozaki, Ken Fukuda IEEE Access, Vol.11, pp.142030-142042, 2023 Code1 / Code2 / Dataset / DOI: https://doi.org/10.1109/ACCESS.2023.3341029 (open access) This study introduces RDF-star2Vec, a novel KGE model designed for RDF-star graphs, which are (recursive) hyper-relational knowledge graphs. In addition, we provide a dataset and a benchmarking framework for data mining tasks focused on complex RDF-star graphs. |
|
13.
Shusaku Egami, Takanori Ugai, Mikiko Oono, Koji Kitamura, Ken Fukuda IEEE Access, Vol.11, pp.23857-23873, 2023 Code / Dataset / DOI: https://doi.org/10.1109/ACCESS.2023.3253807 (open access) We proposed the VirtualHome2KG framework to generate synthetic KGs of daily life activities in virtual space. We also demonstrated the utility and potential of the VirtualHome2KG through several use cases, including the analysis of daily activities by querying, embedding, and clustering, and fall risk detection among older adults based on expert knowledge. |
|
12.
Shusaku Egami, Yasunori Yamamoto, Ikki Ohmukai, Takashi Okumura PLOS ONE, Vol.18, No.3: e0282291, 2023 data / project page / DOI: https://doi.org/10.1371/journal.pone.0282291 (open access) We constructed an ontology, CIRO, which can infer the risk of COVID-19 infection from the action history for the actual operation of tracking and screening of close contacts at public health centers. |
|
11.
Shusaku Egami, Takahiro Kawamura, Kouji Kozaki, Akihiko Ohsuga Data Intelligence, Vol.4, No.1, pp.88-111, 2022 DOI: https://doi.org/10.1162/dint_a_00113 (open access) We extracted urban problem causality from various documents and structured the data as a KG. Then we detected vicious cycles and root problems using SPARQL and SWRL. Furthermore, urban-problem experts evaluated the extracted causal relations. |
|
10.
Yuto Tsukagoshi, Shusaku Egami, Yuichi Sei, Yasuyuki Tahara, Akihiko Ohsuga IEEJ Transactions on Electronics, Information and Systems, Vol.141, No.11, pp.1222-1233, 2021 (in Japanese) DOI: https://doi.org/10.1541/ieejeiss.141.1222 We collected unstructured data from a university campus and integrated it as a knowledge graph based on the proposed ontology. |
|
9.
Shusaku Egami, Xiaodong Lu, Tadashi Koga, Yasuto Sumiya Transactions of the Japanese Society for Artificial Intelligence, Vol.36, No.1, pp.WI2-F_1-12, 2021.1 (in Japanese) project page / DOI: https://doi.org/10.1527/tjsai.36-1_WI2-F (open access) We developed a reference ontology that enables common situational awareness of spatiotemporal concepts for semantic interoperability in air traffic information management. |
|
8.
Takahiro Kawamura, Shusaku Egami IEEE Transactions on Engineering Management, Vol.68, No.5, 2021 project page / DOI: https://doi.org/10.1109/TEM.2019.2946886 (open access) We proposed a method for creating word and paragraph vectors corresponding to bilingual textual information in the same multidimensional space, aiming to construct a bilingual map of science. |
|
7.
Shusaku Egami, Takahiro Kawamura, Kouji Kozaki, Akihiko Ohsuga International Journal of Smart Computing and Artificial Intelligence, Vol.3, No.1, pp.71-86, 2019 DOI: https://doi.org/10.52731/ijscai.v3.i1.321 (open access) We extracted causal relations using natural language processing and crowdsourcing to construct urban problem Linked Data. |
|
6.
Takahiro Kawamura, Katsutaro Watanabe, Naoya Matsumoto, Shusaku Egami, Mari Jibu Scientometrics, Vol.116, pp.941-958, 2022 project page / DOI: https://doi.org/10.1007/s11192-018-2783-x (open access) We proposed a new content-based method of locating research projects in a multi-dimensional space using the word/paragraph embedding techniques. |
|
5.
Shusaku Egami, Takahiro Kawamura, Akihiko Ohsuga IEICE Transactions on Information and Systems, Vol.E101-D, No.1, pp.116-129, 2018 DOI: https://doi.org/10.1587/transinf.2017SWP0010 (open access) We complemented temporal and spatial missing data of the Linked Open Data (LOD) of the problem of illegally parked bicycles using bayesian networks and computational fluid dynamics. |
|
4.
Shusaku Egami, Takahiro Kawamura, Akihiko Ohsuga International Journal of Smart Computing and Artificial Intelligence, Vol.1, No.2, pp.77-93, 2017 DOI: https://doi.org/10.52731/ijscai.v1.i2.99 (open access) We proposed a schema of illegally parked bicycle LOD (IPBLOD) and a methodology of designing LOD schema. |
|
3.
Shusaku Egami, Takahiro Kawamura, Akihiko Ohsuga Transactions of the Japanese Society for Artificial Intelligence, Vol.31, No.6, pp.AI30-K_1-12, 2016 (in Japanese) DOI: https://doi.org/10.1527/tjsai.AI30-K (open access) We purposed eco-cycle for solving illegally parked bicycles using linked open data. |
|
2.
Shusaku Egami, Takahiro Kawamura, Yuichi Sei, Yasuyuki Tahara, Akihiko Ohsuga Transactions on Large-Scale Data and Knowledge-Centered Systems XXVII, Springer LNCS, Vol.9860, pp.129-141, 2016 DOI: https://doi.org/10.1007/978-3-662-53416-8_8 We built an ecosystem that generates Open Urban Data in Link Data format while complementing missing attribute values. |
|
1.
Shusaku Egami, Takahiro Kawamura, Akihiro Fujii, Akihiko Ohsuga IEICE Transactions on Information and Systems, Vol.J98-D, No.6, pp.992-1004, 2015 (in Japanese) CRES: http://id.nii.ac.jp/1438/00009010/ (open access) DOI: https://doi.org/10.1007/978-3-662-53416-8_8 We constructed a linked open data of industrial parts (screw LOD) to realize a business support agent that applies the screw LOD to the bill of materials (BOM). |