Empty response from a trained custom NER ML Model

News Knowledge Base
, , ,
1

Empty result using Custom Named Entity Recognition model as ML skill in Studio, after being trained with data.

After training Custom Named Entity Recognition ML model with data, while using the ML skill in Studio, it is giving empty response as below.

Same result is seen while using the ML skill as endpoint with API key.

Resolution:

  • Download the train pipeline logs and check for the accuracy.

For better results, the test score should be greater than 0.8.

To achieve such score,

  1. The number of epochs for which pipeline runs can be increased in pipeline parameters before starting the pipeline. By default, number of epochs of this ML model is configured to 5.

  1. More samples can be added in the train dataset.