But I made some minor changes to the original repo, in order to make it compatible with Chinese BERT
I deployed the model on HuggingFace Space, link:https://huggingface.co/spaces/diagaiwei/ir_chinese_medqa
我把模型部署到了hf space,链接 https://huggingface.co/spaces/diagaiwei/ir_chinese_medqa
- it requires 3 files, queries.tsv, doc.tsv, triplets.jsonl
-
the content of queries.tsv should be like this ( header is not needed in the tsv)
query_1_id \t 高血压吃什么药 ?
query_2_id \t 糖尿病吃什么药 -
the content of doc.tsv should be like this ( header is not needed in the tsv):
doc_1_id \t 高血压应该吃药A
doc_2_id \t 糖尿病应该吃药B -
each line in the triplets file should be like this: query_id, pos_doc_id,neg_doc_id, for example:
[query_1_id, doc_1_id, doc_2_id]
[query_1_id, doc_1_id, doc_3_id]
[... ... .... ]
-