Homeland

Published: 4 February 2026| Version 2 | DOI: 10.17632/p8cc46znd4.2
Contributor:
Meiqi Wang

Description

The Homeland dataset is a game-based personalized dialogue dataset built to overcome the limitations of existing counterparts, such as narrow persona types and shallow dialogues. It features self-created global personas (18 structured/unstructured types with unique PIDs) from subjective and objective dimensions, with dialogues collected via random free interaction without mandatory persona use. The dataset includes high-reliability manual local persona annotations (92.02% inter-annotation consistency), plus multi-dimensional labels for emotion, topic and dialogue quality. Comparative analysis shows it has more dialogue turns, a realistic 44.00% local persona occurrence ratio, and tends to in-depth discussion of core persona types like identity, with scalable collection potential for future expansion. Here we provide some data samples.

Files

Categories

Natural Language Processing, Dialogue

Licence