A collection of many Japanese roleplaying datasets reformatted the same way.
Each row contains a conversation as well as the source dataset, its dataset config, and split.
Example:
{
"conversations": [
{
"from": "system",
"value": "次に示す世界観やキャラクター設定を参照し、それを元にロールプレイのような対話をしてください。..."
},
{
"from": "human",
"value": "勇人「や、やめてください...魔族の女王...」（顔を赤らめながら、必死に目をそらす）"
},
{
"from": "gpt",
"value": "リリス「ふふ、そんなに抵抗しても無駄よ。あなたの体は正直ね」..."
},
...… See the full description on the dataset page:
https://huggingface.co/datasets/agentlans/Aratako-Japanese-Roleplay.
