AWARE

AWARE 数据集的同行评审论文发表在 ASEW 2021，可通过以下方式访问：http://doi.org/10.1109/ASEW52652.2021.00049。使用 AWARE 数据集时请引用本文。 基于方面的情绪分析 (ABSA) 旨在识别关于特定方面的意见（情绪）。由于缺少注释以支持 ABSA 任务的智能手机应用程序评论数据集，我们提出了 AWARE：ABSA 应用程序评论仓库。 AWARE 包含来自三个不同领域（生产力、社交网络和游戏）的应用评论，因为每个领域都有其不同的功能和受众。每个句子都标注了三个标签，如下所示： 方面术语：存在于句子中的术语，描述了应用程序的一个方面，由情感表达。 “N/A”的术语值意味着该术语没有在句子中明确提及。 方面类别：预定义的一组特定于域的类别，代表应用程序的一个方面（例如，安全性、可用性等）。 情绪：正面或负面。 注意：游戏域不包含方面术语。 我们提供了来自三个领域的 11323 个句子的综合数据集，其中每个句子都附加了一个布尔值注释，表明该句子是否表达了正面/负面意见。此外，我们提供了三个独立的数据集，每个域一个，只包含表达意见的句子。名为“AWARE_metadata.csv”的文件包含数据集列的描述。 如何使用 AWARE？ 我们设计了 AWARE，使其可用于服务于各种任务。任务可以是但不限于： 情绪分析。 方面术语提取。 方面类别分类。 方面情绪分析。 显式/隐式方面术语分类。 意见/非意见分类。 此外，研究人员可以试验和调查不同领域对用户反馈的影响。The peer-reviewed paper for the AWARE dataset was published at ASEW 2021, and can be accessed via: http://doi.org/10.1109/ASEW52652.2021.00049. Please cite this paper when using the AWARE dataset. Aspect-Based Sentiment Analysis (ABSA) aims to identify opinions (sentiments) toward specific aspects. Owing to the lack of annotated smartphone app review datasets for ABSA tasks, we propose AWARE: ABSA App Review Repository. AWARE contains app reviews from three distinct domains: productivity, social networking, and gaming, as each domain has unique features and target audiences. Each sentence is annotated with three labels, as detailed below: - Aspect Term: A term present in the sentence that describes an aspect of the app, toward which sentiment is expressed. A term value of "N/A" indicates that no term is explicitly mentioned in the sentence. - Aspect Category: A predefined set of domain-specific categories that represent an aspect of the app (e.g., security, usability, etc.). - Sentiment: Positive or negative. Note: The gaming domain does not include aspect terms. We provide a comprehensive dataset of 11,323 sentences across the three domains, with each sentence attached with a boolean annotation indicating whether the sentence expresses positive/negative opinions. Additionally, we offer three separate datasets, one for each domain, that only contain opinion-expressing sentences. The file named "AWARE_metadata.csv" contains descriptions of the dataset columns. How to use AWARE? We designed AWARE to support a wide range of tasks. Tasks include but are not limited to: - Sentiment Analysis - Aspect Term Extraction - Aspect Category Classification - Aspect-Based Sentiment Analysis - Explicit/Implicit Aspect Term Classification - Opinion/Non-opinion Classification In addition, researchers can experiment with and investigate the impact of different domains on user feedback.

搜集汇总

背景与挑战

背景概述

AWARE是一个基于方面的情感分析数据集，专门针对智能手机应用评论，涵盖生产力、社交网络和游戏三个领域，包含11323个标注句子。该数据集支持多种自然语言处理任务，如情感分析和方面术语提取，并已通过同行评审论文发布。