Switchboard Coherence (SWBD-Coh) corpus
收藏资源简介：
Switchboard Coherence (SWBD-Coh) corpus是一个开放领域对话资源，包含根据上下文（对话历史）对潜在的下一轮对话候选进行人类一致性评级的注释。评级尺度为：1 = 不一致，2 = 不确定是否合适，3 = 一致。该资源由1000个上下文与7个下一轮对话候选配对组成，共7000个注释的上下文/候选对。
The Switchboard Coherence (SWBD-Coh) corpus is an open-domain dialogue resource that includes annotations of human coherence ratings for potential next-turn dialogue candidates based on context (dialogue history). The rating scale is as follows: 1 = incoherent, 2 = uncertain if appropriate, 3 = coherent. This resource consists of 1000 contexts paired with 7 next-turn dialogue candidates, totaling 7000 annotated context/candidate pairs.
Switchboard Coherence (SWBD-Coh) corpus
数据集概述
- 名称: Switchboard Coherence corpus (SWBD-Coh)
- 描述: 一个开放领域的对话数据集，包含1000个对话上下文，每个上下文有7个潜在的下一轮对话候选，总计7000个上下文/候选对。
- 注释: 每个候选对话根据与上下文的连贯性进行评分，评分标准为：1 = 不连贯，2 = 不确定是否合适，3 = 连贯。
- 注释过程: 通过Amazon Mechanical Turk进行，每个对话由5名通过预选测试的工人进行评分。
数据结构
-
JSON格式: json { "<example_id>": { "candidates": [ { "cand_type": <cand_type>, "avg_score": <avg_score> } ], "context": [ { "speaker": <speaker>, "turn": <turn> } ], "info": { "dialog_id": <dialog_id>, "time_info": [ { "ann_id": <ann_id>, "time": <time> } ], "candidates_info": [ { "cand_type": <cand_type>, "turn_idx": <turn_idx>, "dialog_id": <dialog_id>, "annotators_info": [ { "ann_id": <ann_id>, "score": <score> } ] } ] } } }
- 字段解释:
- example_id: 注释示例的ID。
- candidates: 候选对话信息，包括类型和平均评分。
- context: 对话历史，包含说话者和对话内容。
- info: 额外信息，包括对话ID、注释者信息和时间信息。
- 字段解释:
引用信息
- 引用格式: bibtex @inproceedings{cervone2020dialogue, title={Is this Dialogue Coherent? Learning from Dialogue Acts and Entities}, author={Cervone, Alessandra and Riccardi, Giuseppe}, booktitle={Proceedings of the 21st Annual SIGdial Meeting on Discourse and Dialogue}, year={2020} }
