This master's thesis from 2010 analyzes user comments on videos using the tf-idf method to interactively suggest related videos. The author treats all comments on a single video as a document and calculates feature vectors for each document. A user's "comment tendency vector" is synthesized from videos they've watched. Related videos are identified by calculating similarity between this vector and feature vectors of other videos. The author has one published paper on visualizing research outputs using ExtJS and Google Chart.
1. 複合情報学専攻 修士論文 (2010 年 2 月 10 日)
ユーザコメントの tf-idf 法による分析を用いた
インタラクティブな関連動画の提示
複雑系工学講座 調和系工学研究室 修士 2 年 江端佑介
Interactive Suggestion of Related Videos
by Analyzing Users' Comments Based on Tf-idf Scheme
Research Group of Complex Systems Engineering
Laboratory of Harmonious Systems Engineering
MC2 Yusuke Ebata
Abstract: These days video sharing site such as YouTube or NicoNico-Video is being popular. With
that, studies using comments of videos are also paid more and more attention. In the information which
we can give to videos such as a title and a tag, the comments are the only information which is given by
the user who watched a video. So I can say that users' comments have some information about the
contents of a video. In this study, I show the related videos for which comments of videos was used.
Specifically, I regarded all comments of one video as one document, and calculated feature vector on
every document. Furthermore, I synthesized feature vector of comments of some videos which
a user watched, and defined that vector as a comment tendency vector. And I indicated the
related video by calculating the similarity between a comment tendency vector and feature
vector of a movie.
研究業績(査読付き学術論文,国際会議講演論文,国内講演論文等):
1. 江端佑介,川村秀憲,鈴木恵二,” 研究業績の視覚化に向けた ExtJS と Google Chart による実装”, 情報処理北
海道シンポジウム 2009,平成 21 年 10 月.