I created a custom ML using the Data manager and labeled each field as it should be labelled and then I run a training, evaluation and full pipeline which went successful.
The model is working fine if the number of input files are 5,10,15 etc.
But when I input a bulk of files e.g. 100 or more. then the fields extracted by the ML extractor are not very satisfying.
like even if the name is clear in the document it still misspell the words and every field when the files are run as a bulk.
Can you suggest some way or hint that why is it happening in the case of bulk input files but not in the case of single input file ?