KSE-RESEARCH-Group/ukrainian-dialogs-constructiveness

乌克兰在线话语：双轴建设性数据集是一个专家标注的乌克兰语对话数据集，来源于Telegram和Ukrayinska Pravda论坛。该数据集基于两个正交轴——关系行为（RC）和实质性贡献（SC），对对话的建设性进行了六个有序维度的评分。数据集包含两个部分：gold部分有300条数据，每条数据由三位校准专家标注；main部分有1,071条数据，每条数据由两位标注者标注，其中189条数据由第三位标注者进行验证。每条数据包括一条目标消息及其前九条直接回复链的上下文，提供了语用背景。文本字段以原始西里尔字母形式和拉丁字母转写形式呈现。数据集适用于有序分类、对话健康分析和语用上下文研究等任务。

The Ukrainian Online Discourse: A Two-Axis Constructiveness Dataset is an expert-annotated collection of Ukrainian-language dialogues from Telegram and the Ukrayinska Pravda Forum. It scores dialogic constructiveness along six ordinal dimensions organized into two orthogonal axes: Relational Conduct (RC) and Substantive Contribution (SC). The dataset includes two splits: gold with 300 rows, each annotated by three calibrated experts, and main with 1,071 rows, each annotated by two annotators, with a third annotator verifying a subset of 189 items. Each item consists of a target message and up to nine preceding messages from the same direct-reply chain, providing pragmatic context. Text fields are presented in both original Cyrillic and Latin-script transliteration. The dataset is suitable for tasks such as ordinal classification, conversational health analysis, and pragmatic context studies.