Bilingual (English + Korean) training data for an LSTM-based seq2seq debate
chatbot. Each record is a (topic, input_context, target_output) triple
plus precomputed encoder_input / decoder_input / decoder_target ready
for seq2seq training.
{
"id": "ibm_argq_30k_8b4b12caccad",
"lang": "en",
"source": "ibm_argq_30k",
"is_synthetic": false,
"input_stance": "pro",
"target_stance": "con",
"topic": "We should abandon marriage"… See the full description on the dataset page:
https://huggingface.co/datasets/ada-flo/nlp-hack-debate.
