Home > About Us > Technology > Technical Report > Technical Report No.23 2014

Modeling Sentiment Topics with Distant Supervision

The recent popularity of social media has encouraged many individuals to publically express their opinions. These opinions include valuable information about a vast array of topics. This paper proposes a method of automatically extracting sentiment topics from these social media texts.
Topic modeling is a well-known method of extracting topics expressed as the probabilistic distributions of words from a text collection. We introduce an idea based on distant supervision to this topic modeling, and enable an extraction of topics associated with sentiments. By using the proposed method, we show that sentiment topics can be easily extracted with little effort.

Author

  • Yasuhide Miura
    Communication Technology Laboratory, Research & Technology Group
  • Tomoko Ohkuma
    Communication Technology Laboratory, Research & Technology Group
  • Keigo Hattori
    Communication Technology Laboratory, Research & Technology Group
  • Hiroshi Masuichi
    Communication Technology Laboratory, Research & Technology Group