SPACE (Opinion Summarization) Dataset

SPACE is a large-scale opinion summarization benchmark for the evaluation of unsupervised summarizers. SPACE is built on TripAdvisor hotel reviews and includes a training set of approximately 1.1 million reviews for over 11 thousand hotels.

For evaluation, we created a collection of human-written, abstractive opinion summaries for 50 hotels, including high-level general summaries and aspect summaries for six popular aspects: building, cleanliness, food, location, rooms, and service. Every summary is based on 100 input reviews, an order of magnitude increase compared to existing corpora. In total, SPACE contains 1,050 gold standard summaries. You can view the full instructions for out multi-stage annotation procedure here.

SPACE是一项针对无监督摘要评估的大规模观点摘要基准。SPACE构建于TripAdvisor的酒店评论之上，包含约110万条针对超过11,000家酒店的训练集。为了评估，我们为50家酒店创建了一组由人类编写的、抽象性观点摘要，包括对六个流行方面的概述性摘要和方面摘要：建筑、清洁度、食物、位置、客房和服务。每个摘要基于100条输入评论，相较于现有语料库增加了数量级。总体而言，SPACE包含了1,050条黄金标准摘要。您可以在此处查看我们多阶段注释流程的完整说明。