Presentation | 2017-11-24 Analyses of Neural Language Model and Its Application to Transformation of Individuality in Speech Tatsuya Takeuchi, Masafumi Hagiwara, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this research, we aim to convert the output indirectly by changing the internal state of the neural language model used in a dialogue system, and to change the individuality of the dialogue system. In the process of dimension reduction and conversion into an embedding vector, we acquire the value of the generated embedding vector, and analyze the relationship with the output sentence. At the same time, we propose a system which converts specific words in a sentence by using the feature value obtained. In the proposed system, we used the encoder - decoder model to obtain the differences between the embedding vectors of the input sentences when only specific words were converted, and analyzed them. As a result, characteristic values were confirmed in the vector difference. In addition, we transformed the embedding vector in the middle layer using the difference, and it was shown that some changes occurred in the output sentences. Evaluation experiments confirmed that it is possible to change the individuality of the dialog system using the conversion. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Neural Language Model / Transformation of Individuality in Speech / Long short-term memory |
Paper # | NC2017-29 |
Date of Issue | 2017-11-17 (NC) |
Conference Information | |
Committee | MBE / NC |
---|---|
Conference Date | 2017/11/24(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Tohoku University |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | BCI/BMI, generall |
Chair | Kazuki Nakajima(Univ. of Toyama) / Masafumi Hagiwara(Keio Univ.) |
Vice Chair | Masaki Kyoso(TCU) / Yutaka Hirata(Chubu Univ.) |
Secretary | Masaki Kyoso(Toyama Pref. Univ.) / Yutaka Hirata(Kindai Univ.) |
Assistant | Kim Juhyon(Univ. of Toyama) / Takumi Kobayashi(YNU) / Yoshihisa Shinozawa(Keio Univ.) / Keiichiro Inagaki(Chubu Univ.) |
Paper Information | |
Registration To | Technical Committee on ME and Bio Cybernetics / Technical Committee on Neurocomputing |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Analyses of Neural Language Model and Its Application to Transformation of Individuality in Speech |
Sub Title (in English) | |
Keyword(1) | Neural Language Model |
Keyword(2) | Transformation of Individuality in Speech |
Keyword(3) | Long short-term memory |
1st Author's Name | Tatsuya Takeuchi |
1st Author's Affiliation | Keio University(Keio Univ.) |
2nd Author's Name | Masafumi Hagiwara |
2nd Author's Affiliation | Keio University(Keio Univ.) |
Date | 2017-11-24 |
Paper # | NC2017-29 |
Volume (vol) | vol.117 |
Number (no) | NC-325 |
Page | pp.pp.13-18(NC), |
#Pages | 6 |
Date of Issue | 2017-11-17 (NC) |