Research Topics

We propose approaches based on probability statistics, signal processing, and machine learning to analyze and support intellectual creation activities.

Multimedia Retrieval

We are conducting sentiment analysis and popularity prediction of multimedia contents. We are also studying classification and retrieval methods for data in specific domains such as map images and fashion images.

  • Classification of map images [Sawada&Katsurai, ICASSP2020]
  • Classification of fashion style [Ikeda+,DEIM2019] [Uemura+, DEIM2019]
  • Popular prediction of recipe images [Sanjo&Katsurai, CIKM2017]
  • Emotion classification [Katsurai&Satoh, ICASSP2016]

Microblog Analysis

We propose a method for automatically assigning emotion scores to words not included in general dictionaries (e.g. pictographs and slang).

  • Emoji emotion score [Kimura&Katsurai, FAB2017(ASONAM2017)]
  • Comparison of Emoji between Japanese and Emoji tweets [Kimura&Katsurai, iiWAS2018]
  • Building a slang sentiment dictionary for Nico Nico Douga [Ogura&Katsurai, DEIM2017]

Linked Data

Relating data on the web to each other is a fundamental technique for knowledge discovery and machine learning. We are currently working on the automatic linking of records between different databases.

  • Author matching of academic data [Katsurai&Ohmukai, JCDL2019]
  • Author matching of academic data in different languages [Chikazawa+ DEIM2019]

Researcher Topic Analysis

The system estimates the researcher's expert topic from the text of the article title, keywords, and abstract, and applies it to the search of related researchers and author identification.

  • Construction of a researcher search interface [Tango+, DEIM2020]
  • Topic transition visualization [Nishizawa+, APSIPA ASC 2018]
  • Extraction of domestic researchers' topics [Katsurai+, IEICETrans, 2016]

Social Network Analysis

Networks represent the structure of collaboration in intellectual creative activities and analyze the activity within communities and the differences between communities.

  • Derivation of Research IR Indicators [Araki+, DBSJ, 2018]
  • Analysis of the current state of collaboration within research institutions [Araki+, IEICE Trans, 2017]

Trend Mapping

Knowing people's interests helps in strategy development and marketing. Analyze large amounts of data to extract and visualize prevailing topics in each age group.

  • Visualization of research trends [Katsurai+Ono, Scientometrics, 2019]