2017.10.20
Short term goal: (1)Finish the experiment : build topic Evolution roads. (2) Think how to do use the topic evolution method in Danmaku datasets.
1.Do the experiment.
2.Discuss with Dr. Sun Yunling about the idea of Danmaku comments in her area, and she gives me some suggestions.
2017.10.19
Short term goal: (1)Finish the experiment : build topic Evolution roads. (2) Think how to do use the topic evolution method in Danmaku datasets.
1.Discuss with Prof. He about the idea of topic evolution,and he gives me some suggestions. Dr. Chenqin suggest me have a try using LSTM.
2.Discuss with Prof. Fang about the detail reduction of jiont NMF.
3.Do the experiment.
2017.10.18
Short term goal: (1)Finish the experiment : build topic Evolution roads. (2) Think how to do use the topic evolution method in Danmaku datasets.
1.Read some papers about topic discovery in video topics discovery(Using danmaku).
<Reading the Videos>(AAAI 2016)<Crowdsourced Time-sync Video Tagging using Temporal and Personalized Topic Modeling >(KDD 2014)\Web Video Topic Discovery and Tracking via Bipartite Graph Reinforcement Model(WWW 2008,not damaku datasets).
2.Sumarizes the feasible method.
3.Disscuss with Guowang about the paper the paper of <Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction>.
2017.10.17
Short term goal: (1)Finish the experiment : build topic Evolution roads. (2) Think how to do use the topic evolution method in Danmaku datasets.
1.Analyse some bilibili Danmaku comments using nmf:add jieba tool for Chinese words segmentation,the result seems not good. Dengyong suggest me use LTP tool. I will try tomorrow.
- Sumarizes the feasible method of Danmaku .
3.Talk with Jiangjian (the student of prof.lv) about the topic visualization.
2017.10.16
Short term goal: (1)Finish the experiment : build topic Evolution roads. (2) Think how to do use the topic evolution method in Danmaku datasets.
- Discussing what kind of Danmaku comments to crawl with Songyang and Wang tingting. She will help us crawl the comments from Youku website.
2.Analyse some bilibili Danmaku comments using both topic model and textrank model.Results show that the topic is not separated clearly. I compared cartoon comments and TV shows, the words have a tough connection between video and comments.
网友评论