Sports Feedback
收藏资源简介：
本研究引入了一个创新的合成数据集——Sports Feedback，该数据集由圣玛丽大学的研究人员创建，包含了体育赛事志愿者的反馈信息。这个数据集特别之处在于其领域特定术语和大量的隐性方面，为Aspect-Based Sentiment Analysis任务带来了独特的挑战。该数据集共有480份文档，平均每份文档222个字符，其中75%由GPT-4生成，25%由Gemini 1.0 Ultra生成，以引入内容多样性。数据集的创建过程包括三个步骤：首先由LLM生成初步注释草稿，然后由志愿者从中选择更合适的草稿，最后由专家进行修订和完善。该数据集的35%内容含有隐性方面，为评估大型语言模型在提取隐性方面的能力提供了良好的测试平台。
This study introduces an innovative synthetic dataset named Sports Feedback, developed by researchers from Saint Mary's University, which contains feedback information collected from sports event volunteers. What distinguishes this dataset is its domain-specific terminology and a large number of implicit aspects, posing unique challenges for the Aspect-Based Sentiment Analysis (ABSA) task. The dataset consists of 480 documents, with an average length of 222 characters per document. Specifically, 75% of the content is generated by GPT-4 and the remaining 25% by Gemini 1.0 Ultra, to introduce content diversity. The dataset creation process includes three steps: first, generating preliminary annotation drafts via Large Language Models (LLMs), then having volunteers select more appropriate drafts from the generated outputs, and finally revising and polishing the selected drafts with the guidance of domain experts. Thirty-five percent of the content in this dataset contains implicit aspects, making it an excellent testbed for evaluating the ability of Large Language Models (LLMs) to extract implicit aspects.
数据集概述
数据集描述
该数据集（dataset.json）专为基于方面的情感分析设计，包含文本数据及针对文本中提及的各个方面的情感标注。
数据字段说明
text
- 类型:
string - 描述: 包含主要的文本内容，代表被分析或处理的文本数据。示例包括：
- "Ive always believed in the importance of inclusive sporting events like NBMSG, but the execution this year was lacking. From the get-go, information was scarce, leaving many of us unsure of our roles and responsibilities."
- "The food stalls, despite some limitations, did a fantastic job of offering a taste of home to many attendees. It was a nice touch that added to the overall welcoming atmosphere of the event."
- "Volunteering gave me a sense of purpose. Its cool to be part of something thats all about celebrating diversity and bringing people together."
aspects
- 类型:
dict - 描述: 包含一个字典，其中每个键值对代表文本中的一个特定方面及其相关的情感极性。每个方面是文本中提到的特征或属性，值表示对该方面的情感（如Positive、Negative）。示例包括：
{"inclusivity": "Positive", "execution": "Negative", "information": "Negative"}{"food stalls": "Positive", "atmosphere": "Positive"}{"volunteering": "Positive", "sense of purpose": "Positive", "diversity": "Positive", "bringing people together": "Positive"}
is_train
- 类型:
boolean - 描述: 指示该行数据是否属于训练集或测试集。值为
True表示属于训练集，False表示属于测试集。
数据示例
以下是数据集中的一些示例条目：
json { "text": "Ive always believed in the importance of inclusive sporting events like NBMSG, but the execution this year was lacking. From the get-go, information was scarce, leaving many of us unsure of our roles and responsibilities.", "aspects": { "inclusivity": "Positive", "execution": "Negative", "information": "Negative" }, "is_train": false }, { "text": "The food stalls, despite some limitations, did a fantastic job of offering a taste of home to many attendees. It was a nice touch that added to the overall welcoming atmosphere of the event.", "aspects": { "food stalls": "Positive", "atmosphere": "Positive" }, "is_train": false }, { "text": "Volunteering gave me a sense of purpose. Its cool to be part of something thats all about celebrating diversity and bringing people together.", "aspects": { "volunteering": "Positive", "sense of purpose": "Positive", "diversity": "Positive", "bringing people together": "Positive" }, "is_train": false }
