KOMODIS-Dataset

KOMODIS-Dataset是由波茨坦大学计算语言学系与德国大众汽车公司合作创建的一个增强对话数据集，专注于电影讨论领域。该数据集通过亚马逊Mechanical Turk平台收集，包含7519条对话，每条对话都基于预先指定的电影事实和观点。数据集的创建过程涉及为每个对话生成独特的知识和观点组合，确保对话内容的丰富性和相关性。KOMODIS-Dataset的应用领域主要集中在非目标驱动对话的神经模型训练，旨在解决对话模型中知识和观点一致性的问题，提高对话的自然度和知识性。

KOMODIS-Dataset is an enhanced dialogue dataset co-created by the Department of Computational Linguistics, University of Potsdam and Volkswagen AG Germany, focusing on the domain of film discussion. This dataset was collected via the Amazon Mechanical Turk platform, containing 7519 dialogues, each based on pre-specified film facts and opinions. The dataset creation process involves generating unique knowledge and opinion combinations for each dialogue, ensuring the richness and relevance of the dialogue content. The primary application scenarios of KOMODIS-Dataset focus on neural model training for non-goal-oriented dialogues, aiming to address the issue of knowledge and opinion consistency in dialogue models and improve the naturalness and informativeness of dialogues.