Cost Clarification and Optimization Options for Custom NER Project

Question

Dear Azure Pricing Team,

I am currently working on a custom Named Entity Recognition (NER) project and have been using Azure Language Studio for training and deployment. However, I’ve noticed that my costs have been higher than expected, and I would like to understand the billing structure better and explore potential optimization options.

Here are my specific questions:

Character Limit per Record in Language Studio: Could you clarify the character limit for each "text record" in the Standard pricing tier? I noticed that while processing approximately 200 records, my charges were around $50, which seemed high compared to the expected rate of $5 per 1,000 records.

Options to Reduce Costs:

• Are there any configurations within Azure Language Studio that could help reduce costs? For instance, does splitting larger documents into smaller records impact the pricing?
• Can you recommend any specific pricing tiers, configurations, or alternative Azure services, such as Azure Machine Learning, that might be more cost-effective for custom NER models? Comparison with Azure Machine Learning (AML):
  • Are there any configurations within Azure Language Studio that could help reduce costs? For instance, does splitting larger documents into smaller records impact the pricing?
    • Can you recommend any specific pricing tiers, configurations, or alternative Azure services, such as Azure Machine Learning, that might be more cost-effective for custom NER models?
    Comparison with Azure Machine Learning (AML):

        - Based on your experience, would shifting this project to Azure Machine Learning help manage costs better for a high-volume, document-heavy NER workload?
        
        
           - If I transition to AML, are there any particular configurations or cost-saving measures (e.g., using spot instances, scaling options) that you recommend to maintain performance while controlling expenses?
    

Answer 1

Hello Dinnemidi Ananda Kumar,

Welcome to Microsoft Q&A! Thanks for posting the question.

Here are the responses for your queries:

Character Limit per Record in Language Studio

The character limit for each "text record" in the Standard pricing tier of Azure Language Studio is 1,000 characters. This means that if your documents exceed this limit, they are counted as multiple records, which could explain the higher costs you're seeing.

For more detailed information, you can refer to the official documentation on data limits for Azure Language Service Data limits for Language service features and Pricing details.

Options to Reduce Costs

There are several configurations within Azure Language Studio that can help reduce costs:

Splitting larger documents into smaller records can indeed help manage costs more effectively. so, breaking down larger documents ensures you're not exceeding this limit unnecessarily. Choosing the right pricing tier and configurations can significantly impact your costs. The Standard pricing tier offers different rates based on the volume of text records processed, please refer this document Data limits for Language service features.

Based on your requirement you can choose and for high-volume, document-heavy workloads, AML might offer more flexibility and potentially lower costs.

Hope this helps. Do let us know if you any further queries.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful.

Thank you!