We proudly presented our work on social relationship extraction using pedometers (e.g. FitBit) on
The 19th International Symposium on Wearable Computers (ISWC 2015) held on September 2015, Osaka, Japan.
We have presented at the following sessions.
9/11 Early Afternoon 14:30 – 15:50, Hall A
ISWC/Environmental Sensing Systems
Fine-grained Social Relationship Extraction from Real Activity Data under Coarse Supervision
Kota Tsubouchi, Osamu Saisho, Junichi Sato, Seira Araki, Masamichi Shimosaka
(Yahoo! JAPAN Research / The University of Tokyo / Tokyo Institute of Technology)
Understanding social relationships plays an important role in smooth information sharing and project management. In this research, we present a novel approach to extract fine-grained social relationships from coarse supervised data utilizing multiple instance learning (MIL). The experimental results show that our approach with coarse supervision data achieves more precise detection than the existing unsupervised approach, and can extract social relationships with detailed information such as time when they are together even though such information is not included in questionnaires.
The (open accessible) paper is available on acm.org