---
task_categories:
- text-generation
language:
- en
size_categories:
- 1K<n<10K
---
上述九个数据集为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。

#### 以acos数据集中抽取的jsonl文件一条数据举例：

```
{
    "task_type": "generation",
    "dataset": "acos", 
    "input": ["the computer has difficulty switching between tablet and computer ."], 
    "output": "[['computer', 'laptop usability', 'negative', 'difficulty']]",
    "situation": "none", 
    "label": "", 
    "extra": "", 
    "instruction": "    
        Task: Extracting aspect terms and their corresponding aspect categories, sentiment polarities, and opinion words. 
        Input: A sentence
        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.
        Example:  
            Sentence: \"Also it's not a true SSD drive in there but eMMC, which makes a difference.\"  
            Output: [['SSD drive', 'hard_disc operation_performance', 'negative', 'NULL']]' 
    "
}
```
> 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，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。