deniseiras/ABSA_beer

该数据集包含多个子集，专注于葡萄牙语啤酒评论的方面级情感分析（ABSA）。主要数据集包括：1. Reviews Main（步骤3生成）：包含59,982条相关啤酒评论，支持评论分析、定量啤酒属性和评论相关信息。2. Reviews Sample（步骤4.1生成）：作为Reviews Main的子集，包含108条评论，适用于ABSA方法的基准测试和验证。3. ABSA Gold（步骤4生成）：手动标注的数据集，基于Reviews Sample，包含1710个标注的啤酒特性（BC），涵盖方面、类别和情感，作为基于提示方法的黄金标准。4. ABSA Main（步骤4生成）：包含880,373条记录，支持从消费者评论中大规模提取见解，涵盖方面、类别和情感。5. ABSA Final（步骤5生成）：包含基本信息，如评论内容、日期、啤酒风格、评分、方面、类别、情感和年份，用于结果提取和时序分析。整体数据集用于研究大型语言模型在无监督ABSA中的应用，聚焦巴西啤酒的消费者评论，支持市场情报和产品开发。This repository contains multiple datasets for unsupervised Aspect-Based Sentiment Analysis (ABSA) on Portuguese beer reviews. Key datasets include: 1. Reviews Main (step_3_reviews_main.csv): Comprises 59,982 relevant beer reviews, supporting analysis of review comments, quantitative beer attributes, and review-related information. 2. Reviews Sample (step_4_1_reviews_sample.csv): A subset of Reviews Main with 108 reviews, suitable for benchmarking and validating ABSA methodologies. 3. ABSA Gold (step_4_ABSA_Gold.csv): A manually annotated dataset derived from Reviews Sample, containing 1710 annotated Beer Characteristics (BC) with aspect, category, and sentiment, serving as a gold standard for prompt-based evaluation. 4. ABSA Main (step_4_ABSA_main.csv): Comprises 880,373 records for large-scale analysis and insight extraction from consumer reviews, with columns for aspect, category, and sentiment. 5. ABSA Final (step_5_ABSA-FINAL.csv): Includes essential information such as review_comment, review_datetime, beer_style, ratings, aspect, category, sentiment, and year for result extraction and temporal analysis. The datasets are used to study LLM applications in ABSA for Brazilian beer consumer reviews, enabling market intelligence and product development.