| Presentation | 2015-12-17 Multimodal Learning of Object Concepts and Word Meanings by Robots Joe Nishihara, Tatsuya Aoki, Tomoaki Nakamura, Takayuki Nagai, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In previous works, we have shown that robots can learn concepts and word meanings by categorizing multimodal information such as visual, audio, haptic, and word information, that is acquired from their physical embodiments. However, the experiment settings are limited to small scale of objects. In this study, we develop an unsupervised online learning framework to realize long-term online concept and word learning, based on online Multimodal Hierarchical Dirichlet Process (oMHDP). Thanks to oMHDP, the robot can make online estimation of the number of concepts. Another important property of the proposed framework is to learn multimodal concepts and the language model simultaneously. This mutual learning framework of concepts and language improves speech recognition and multimodal categorization performance each other. We are implementing this framework to the robot and conducting experiment of mutual concept learning through interaction with human for a month. In this paper, the effectiveness of proposed method is discussed. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Multimodal categorization / concept learning / learning of word meanings / symbol emergence / language model |
| Paper # | CNR2015-21 |
| Date of Issue | 2015-12-10 (CNR) |
| Conference Information | |
| Committee | CNR |
|---|---|
| Conference Date | 2015/12/17(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Michita Imai(Keio Univ.) |
| Vice Chair | Takahiro Miyashita(ATR) / Masayuki Kanbara(NAIST) |
| Secretary | Takahiro Miyashita(Univ. of Tokyo) / Masayuki Kanbara(Univ. of Tsukuba) |
| Assistant | Kosuke Yoshioka(Panasonic) / Yuka Kobayashi(Toshiba) / Takahiro Matsumoto(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Cloud Network Robotics |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Multimodal Learning of Object Concepts and Word Meanings by Robots |
| Sub Title (in English) | |
| Keyword(1) | Multimodal categorization |
| Keyword(2) | concept learning |
| Keyword(3) | learning of word meanings |
| Keyword(4) | symbol emergence |
| Keyword(5) | language model |
| 1st Author's Name | Joe Nishihara |
| 1st Author's Affiliation | The University of Electro-Communications(UEC) |
| 2nd Author's Name | Tatsuya Aoki |
| 2nd Author's Affiliation | The University of Electro-Communications(UEC) |
| 3rd Author's Name | Tomoaki Nakamura |
| 3rd Author's Affiliation | The University of Electro-Communications(UEC) |
| 4th Author's Name | Takayuki Nagai |
| 4th Author's Affiliation | The University of Electro-Communications(UEC) |
| Date | 2015-12-17 |
| Paper # | CNR2015-21 |
| Volume (vol) | vol.115 |
| Number (no) | CNR-375 |
| Page | pp.pp.29-33(CNR), |
| #Pages | 5 |
| Date of Issue | 2015-12-10 (CNR) |