| Research Abstract |
Online games, each constituting a virtual community with more than ten thousand participants, have recently gained a lot of interests as a new cultural, social, and economical model. From a game operating point of view, it is crucial to promote constructive activities in the virtual world because doing so increases attachment to the game due to having a happy life therein. Accordingly, grasping user trends both inside and outside of the game is necessary.
To grasp user trends, there exist two possible approaches, i.e., action analysis dealing with patterns of playing or moving styles, and communication analysis dealing with chat or bulletin board contents. In this research, the latter approach was adopted. More precisely, targeting at chats inside the game, bulletin boards and blogs outside the game, we proposed methods for "keyword extraction", "discussion visualization", and "community visualization", and we evaluated them using real data.
In keyword extraction, we proposed a new method for detection of influential keywords from texts in bulletin board services. The proposed method considers additionally the relationship between comments and their replies as an index of important words.
In discussion visualization, we proposed a preprocessing method for a tool called KeyGraph that can express the relationship between words. The proposed preprocessing method partitions comments according to their importance. This preprocessing successfully enables grasping of discussions in more detail.
In community visualization, we proposed a method for visualizing user networks by focusing on their common words in chat messages. We evaluated the proposed method with real data from a chat room and chat data at the client side of an online game. We also proposed a method for visualizing users in a bulletin board service by using forest metaphor that can effectively represent important users, number of replies, important words, etc.
|
