Share via

使用FormRecognizer studio 标记的顺序和predict预测结果的顺序不同

dachui 10 Reputation points
2023-05-17T02:26:53.28+00:00

具体如:

base:
“1、低蛋白血症、2、腹内器官损伤,其他的、3、急性阴道炎、4、失血性休克、5、腹膜后血肿、6、弥漫性轴索损伤、7、骶骨骨折、8、开放性胫腓骨干骨折、9、腹腔积血、10、肺水肿、”
predict:
"1、低蛋白血症、2、腹内器官损伤,其他的、3、弥漫性轴索损伤、7、骶骨骨折、8急性阴道炎、4、失血性休克、开放性胫腓骨干骨折、、5、腹膜后血肿、6、9、腹腔积血、10、、动眼创伤性脾破裂"
或者相反,predict的结果顺序错误,base结果正确

或者web端有时候会多出几个“、、、”,虽然pdf文件只有1个“、”

想问下

1、web端https://formrecognizer.appliedai.azure.com/studio 下,标记用的modelId prebuilt-layout 和预测用的模型id相同吗

2、为何会出现上面提到的情况

3、标记或者预测,内部对结果会有特殊处理吗,如果有能不能简单分享下

4、如果想对pdf文件使用ocr, modelID为prebuilt-layout,用得到的ocr.json标记,然后拿来训练,这和使用web端标记然后训练的差别在哪

期待回复

非常感谢^_^

Azure Document Intelligence in Foundry Tools

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.