CCB/cis5300-text-classification
收藏资源简介:
该数据集支持复杂词识别(CWI)任务:给定上下文中的一个词,预测其是否为复杂词(可能对非母语者、儿童或有阅读障碍的人造成困难)或简单词。CWI是词汇简化的第一步——即重写文本使其更易于理解的任务。在简化一个词之前,需要先识别哪些词需要简化。数据集包括一个主配置(包含训练、验证和测试集)和两个用于评估模型泛化能力的领域特定测试集(生物医学和新闻领域)。数据集中的每个词由约10名标注者独立标注,标注结果通过阈值转换为二元标签(简单或复杂)。数据集来源于新闻文章,由母语和非母语的英语使用者标注,旨在识别对非母语者、儿童或有阅读障碍的人可能困难的词。
This dataset supports the Complex Word Identification (CWI) task: given a word in context, predict whether it is complex (likely to be difficult for non-native speakers, children, or people with reading disabilities) or simple. CWI is the first step in lexical simplification — the task of rewriting text to make it more accessible. Before you can simplify a word, you need to identify which words need simplification. The dataset includes a main configuration with train, validation, and test splits, as well as two additional domain-specific test sets (biomedical and news) for evaluating model generalization. Each word in the dataset was independently labeled by approximately 10 annotators, and the labels were binarized using a threshold (simple or complex). The dataset is drawn from news articles and annotated by both native and non-native English speakers to identify words that could be difficult for non-native speakers, children, or people with reading disabilities.




