Last month, Rakuten Institute of Technology sent two people to this year’s Japan Society of Artificial Intelligence conference in Kagoshima to present their research. The event is a major one in the Japanese AI community, with more than 2500 participants for the first time this year, and we were very happy to be able to get together with everyone else in the field to present and share ideas.
The presented paper, named “An entity linking method using multiple Doc2Vec”, tackles the problem of correctly linking words from text with entities in a knowledge base due to things like name variations and entity ambiguity. By using a multi-layered Doc2Vec model, the researchers could show some very promising results in both performance and cost effectiveness. The many visitors crowding the presentations kept the presenters very busy explaining the intricate details of the research, but the strong interest is always a joy to see.
In addition to that, seeing the many other presentations, and having an opportunity to meet with peers and discuss the state of the technology is extremely valuable and interesting. The many new applications coming out of the field, covering every industry from internet services and finance to automobile and real estate, left us more excited than ever for what is to come.
For more info on the event and download the full proceedings, see: