TRANNING Related Isuues .

Ashish Ranjan 20

I am Getting Problem In Fetcing Data , which is formely working Good after tranning new model its accuracy is decreased suddenly . Example Transport Charge is Fecting Previsoly , and after tranning new model it stop fecting transport charges without any changes in transport changes . as of now i have lableled transport charge so no need to alter that , then how can i overcome over this problem ? i am Using Custom Extraction Model Service

kothapally Snigdha 1,635 Reputation points Microsoft External Staff

2025-03-06T23:39:41.46+00:00
Hi Ashish Ranjan,

It sounds like you're experiencing issues with your custom extraction model after retraining, particularly with the extraction of transport charges. Here are some steps you can take to address this problem:

After training your model, check the evaluation details to identify areas where the model is underperforming. Look for the confusion matrix to see if transport charges are being misclassified or not extracted at all.

If transport charges are not being extracted correctly, consider adding more tagged instances of transport charges to your training dataset. This can help the model learn better patterns related to this entity.

Ensure that the dataset used for training is representative of the data you expect in production. If the training data lacks diversity or does not cover the scenarios you encounter, the model may struggle to perform well.

If transport charges are being confused with other entities, it might be worth revisiting your schema design. Ensure that the schema is clear and that entities are distinct enough to prevent overlap.

After making adjustments, test your model with real data to see if the changes have improved performance. This can provide insights into whether the model is now able to extract the transport charges correctly.

I hope these helps you. Thank you!
Ashish Ranjan 20 Reputation points

2025-03-07T05:15:40.36+00:00

Thank you For Helping . Now please assist me where i can find confusion matrix to see if tag are being misclassified or not extracted at all in custom extraction model .
kothapally Snigdha 1,635 Reputation points Microsoft External Staff

2025-03-08T01:10:47.6633333+00:00
Hi Ashish Ranjan,

You can find the confusion matrix for evaluating your custom extraction model by referring to the evaluation metrics for custom named entity recognition models. The confusion matrix will help you compare the expected labels with the ones predicted by your model, allowing you to see if tags are being misclassified or not extracted at all. It provides a holistic view of the model's performance and can help identify entities that are often mistaken for one another.

I hope these helps you. Thank you!

Share via

TRANNING Related Isuues .

Your answer