| Presentation | 2019-12-20 Analysis method and results of solution diversity about topic model Toshio Uchiyama, Tsukasa Hokimoto, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Probabilistic Latent Semantic Analysis and Latent Dirichlet analysis are known as topic models to analyze text data and images. When parameters (= solution) of the topic model are obtained by the optimization algorithm, various solutions are reached due to differences in algorithms, initial values. However, there are many similar solutions, and it is redundant to treat them as different findings. To find non-redundant solution sets, We propose a method to calculate normalized mutual information (NMI) as similarity (inner product) between solutions, apply multidimensional scale method to coordinate values to them, and enable analysis and visualization. Distribution situation of solutions was visualized by experiment using various text data. We also clarified the dependence of representative algorithms (maximum a posterior estimation, collapsed Gibbs sampling, collapsed variational Bayesian inference) and initial values, and confirmed the usefulness of the proposed method. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | topic model / diversity of solutions / Normalized Mutual Information / Multidimensional scaling |
| Paper # | PRMU2019-54 |
| Date of Issue | 2019-12-12 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2019/12/19(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Yoichi Sato(Univ. of Tokyo) |
| Vice Chair | Toru Tamaki(Hiroshima Univ.) / Akisato Kimura(NTT) |
| Secretary | Toru Tamaki(NTT) / Akisato Kimura(OMRON SINICX) |
| Assistant | Yusuke Uchida(DeNA) / Takayoshi Yamashita(Chubu Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Analysis method and results of solution diversity about topic model |
| Sub Title (in English) | |
| Keyword(1) | topic model |
| Keyword(2) | diversity of solutions |
| Keyword(3) | Normalized Mutual Information |
| Keyword(4) | Multidimensional scaling |
| 1st Author's Name | Toshio Uchiyama |
| 1st Author's Affiliation | Hokkaido Information University(HIU) |
| 2nd Author's Name | Tsukasa Hokimoto |
| 2nd Author's Affiliation | Hokkaido Information University(HIU) |
| Date | 2019-12-20 |
| Paper # | PRMU2019-54 |
| Volume (vol) | vol.119 |
| Number (no) | PRMU-347 |
| Page | pp.pp.49-54(PRMU), |
| #Pages | 6 |
| Date of Issue | 2019-12-12 (PRMU) |