update read,e

README.md

@@ -15,8 +15,18 @@ tags:
 
 # Japanese DialoGPT trained with Aozora
 
-## Dataset
-- https://www.aozora.gr.jp/
+**(ja) 青空文庫のセリフで学習した日本語のDialoGPT Smallです**  
+**(en) Japanese DialoGPT Small trained on Aozora Bunko.**
 
-## Base Model
-- https://huggingface.co/rinna/japanese-gpt2-small
+## [Demo](https://huggingface.co/spaces/akiFQC/Japanese_DialoGPT_small_Aozora)
+Demo in this page is not working so well. I recommend you to try it on [Hugging Face Spaces Version](https://huggingface.co/spaces/akiFQC/Japanese_DialoGPT_small_Aozora).
+
+
+## Reference
+- [Aozora-bunko](https://www.aozora.gr.jp/)
+  - Japanese public domain books. 
+  - I extracted the dialogue part from the books and used it as the training data.
+- [japanese-gpt2-small](https://huggingface.co/rinna/japanese-gpt2-small)
+  - Novel Japanese GPT2. I used a small model because of the limitation of GPU memory of my desktop PC(with RTX3060x1) 😢.
+  - I used this model as a pre-trained model.
+- [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536)