Update README.md
Browse files
README.md
CHANGED
|
@@ -1,12 +1,18 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
license: afl-3.0
|
| 3 |
---
|
|
|
|
| 4 |
# Hotel review multi-aspect sentiment classification using T5
|
| 5 |
|
| 6 |
We fine tune a T5 pretrained model to generate multi-aspect sentiment classes. The outputs are whole sentiment, aspect, and aspect+sentiment.
|
| 7 |
|
| 8 |
-
|
| 9 |
|
|
|
|
| 10 |
在右側測試區輸入不同的任務文字
|
| 11 |
|
| 12 |
範例1:
|
|
@@ -17,7 +23,7 @@ We fine tune a T5 pretrained model to generate multi-aspect sentiment classes. T
|
|
| 17 |
整体情绪::位置离逢甲很近
|
| 18 |
|
| 19 |
資料集:
|
| 20 |
-
資料集蒐集自線上訂房網站的顧客留言10050
|
| 21 |
|
| 22 |
輸入與輸出格式:有三個種類任務分別為:
|
| 23 |
|
|
|
|
| 1 |
---
|
| 2 |
+
language:
|
| 3 |
+
- tw
|
| 4 |
+
tags:
|
| 5 |
+
- t5
|
| 6 |
license: afl-3.0
|
| 7 |
---
|
| 8 |
+
|
| 9 |
# Hotel review multi-aspect sentiment classification using T5
|
| 10 |
|
| 11 |
We fine tune a T5 pretrained model to generate multi-aspect sentiment classes. The outputs are whole sentiment, aspect, and aspect+sentiment.
|
| 12 |
|
| 13 |
+
T5情緒面向分類多任務，依據中文簡體孟子T5預訓練模型微調，訓練資料集只有3萬筆，做NLP研究與課程的範例模型用途。
|
| 14 |
|
| 15 |
+
# 如何測試
|
| 16 |
在右側測試區輸入不同的任務文字
|
| 17 |
|
| 18 |
範例1:
|
|
|
|
| 23 |
整体情绪::位置离逢甲很近
|
| 24 |
|
| 25 |
資料集:
|
| 26 |
+
資料集蒐集自線上訂房網站的顧客留言10050筆，整理成3項任務，總筆數變成為3倍，共有30150筆(資料由本實驗室成員YYChang蒐集)。
|
| 27 |
|
| 28 |
輸入與輸出格式:有三個種類任務分別為:
|
| 29 |
|