semeval2015
SemEval_2015_PIT_biencoderunsup-promcse-bert-base-uncased-semeval2015-laptopsbert-large-uncased-semeval2015-restaurantsbert-large-uncased-semeval2015-laptopsunsup-simcse-roberta-large-semeval2015-restaurantsunsup-promcse-bert-base-uncased-semeval2015-restaurantsdistillGPT2-semeval2015-fold0distillGPT2-semeval2015-fold1
SemEval2015Task12These are the datasets for Aspect Based Sentiment Analysis (ABSA), Task 12 of SemEval-2015.semeval-2015
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。
补充：SemEval-2015数据集文件夹中有两个文件夹"laptop"和"restaurant"，其实根据数据集文本的主要围绕主题区分的。抽取的元素方面，laptop和restaurant两文件夹中，数据的抽取元素也不同，laptop抽取的是方面类别和情感极性的二元组，restaurant抽取的是方面术语、方面类别和情感极性的三元组
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… See the full description on the dataset page: https://huggingface.co/datasets/NEUDM/semeval-2015.SemEval2015Task12RawA collection of SemEval2015 specifically designed to aid research in Aspect Based Sentiment Analysis.SemEval2015-RestaurantSemEval2015Task12NLTKA collection of SemEval2015 specifically designed to aid research in Aspect Based Sentiment Analysis.