Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
---
|
5 |
+
|
6 |
+
> 上述数据集为ABSA(Aspect-Based Sentiment Analysis)领域数据集,基本形式为从句子中抽取:方面术语、方面类别(术语类别)、术语在上下文中情感极性以及针对该术语的观点词,不同数据集抽取不同的信息,这点在jsonl文件的“instruction”键中有分别提到,在此我将其改造为了生成任务,需要模型按照一定格式生成抽取结果。
|
7 |
+
|
8 |
+
#### 以acos数据集中抽取的jsonl文件一条数据举例:
|
9 |
+
|
10 |
+
```
|
11 |
+
{
|
12 |
+
"task_type": "generation",
|
13 |
+
"dataset": "acos",
|
14 |
+
"input": ["the computer has difficulty switching between tablet and computer ."],
|
15 |
+
"output": "[['computer', 'laptop usability', 'negative', 'difficulty']]",
|
16 |
+
"situation": "none",
|
17 |
+
"label": "",
|
18 |
+
"extra": "",
|
19 |
+
"instruction": "
|
20 |
+
Task: Extracting aspect terms and their corresponding aspect categories, sentiment polarities, and opinion words.
|
21 |
+
Input: A sentence
|
22 |
+
Output: A list of 4-tuples, where each tuple contains the extracted aspect term, its aspect category, sentiment polarity, and opinion words (if any). Supplement: \"Null\" means that there is no occurrence in the sentence.
|
23 |
+
Example:
|
24 |
+
Sentence: \"Also it's not a true SSD drive in there but eMMC, which makes a difference.\"
|
25 |
+
Output: [['SSD drive', 'hard_disc operation_performance', 'negative', 'NULL']]'
|
26 |
+
"
|
27 |
+
}
|
28 |
+
```
|
29 |
+
> 此处未设置label和extra,在instruction中以如上所示的字符串模板,并给出一个例子进行one-shot,ABSA领域数据集(absa-quad,acos,arts,aste-data-v2,mams,semeval-2014,semeval-2015,semeval-2016,towe)每个数据集对应instruction模板相同,内容有细微不同,且部分数据集存在同一数据集不同数据instruction内容不同的情况。
|