search

baidu / DuReader

Baseline Systems of DuReader Dataset
http://ai.baidu.com/broad/subordinate?dataset=dureader
1.13k stars 308 forks source link

I was implemented the dureader to recognize the story name from a long text but the result is so wired #41

Closed fooSynaptic closed 4 years ago

fooSynaptic commented 5 years ago

I inspected the json format and prepared with labeled answers.

image
fooSynaptic commented 5 years ago

the train file look like

image

and the prepare fns behave like: res = train_json_prepare(raw_paragraph[i], '今天听什么长篇评书？', \ '今天听什么长篇评书？', ans_span[i])

lkliukai commented 5 years ago

I am not sure about the details of your task, and what about the results when you treat your task as an entity tagging task?

fooSynaptic commented 5 years ago

I am not sure about the details of your task, and what about the results when you treat your task as an entity tagging task?

@lkliukai I do not think the NER can replace the reading comprehension task

lkliukai commented 5 years ago

I am not sure about the details of your task, and what about the results when you treat your task as an entity tagging task?

@lkliukai I do not think the NER can replace the reading comprehension task

No, I don't think so. NER is only a part of QA task.

fooSynaptic commented 5 years ago

I am not sure about the details of your task, and what about the results when you treat your task as an entity tagging task?

@lkliukai I do not think the NER can replace the reading comprehension task

No, I don't think so. NER is only a part of QA task.

learnning task is not only result generation method, i understand that you point out NER produce result in a similar form with QA, but their optimize target is very different