Azure AI Language Custom NER vs AML AutoML NER

Question

Hi all,I have seen Custom NER in the Azure AI Language studio as well as the NER under Natural Language Processing in AutomatedML in the AML Studio. What are the key differences between these two and are the underlying algorithms the same? I understand that Custom NER is currently only available as a REST API or within the Lang Studio. Would appreciate any guidance on when to use each service. Thanks in advance!

Answer 1

Hi @Devanshi Thakar

Thank you for using Microsoft Q&A.

I understand that you would like to know the key differences between the Custom NER in Azure AI Language Studio and NER under Natural Language Processing in AutomatedML in the AML Studio. I will be happy to assist you with this.

Custom NER in Azure AI Language Studio and NER under Natural Language Processing in AutomatedML are both services that allow you to extract named entities from text. The choice between Custom NER and NER in AutomatedML depends on factors such as the level of customization and control needed, the expertise of the users, integration requirements, and the complexity of the NER task. Custom NER is suitable for users who require fine-grained control and customization, while NER in AutomatedML is more suitable for users looking for an automated and user-friendly approach to building NER models.

In Azure AI Language Studio NER, users manually create custom projects, tagging data and training models via REST API calls. It offers high customization levels, allowing users to define entity categories and configure project settings. Data must be uploaded in specific formats, supporting text and JSON. Deployment requires manual intervention through REST API calls.

AutoML Text NER in AML Studio automates model training without manual tagging or API calls. It provides less customization, focusing on selecting models based on metrics and hyperparameters. Data must be in ML Table format with labeled columns. Deployment is automatic within Azure ML, allowing direct inference from the deployed model. AutoML Text NER prioritizes ease of use, suitable for users seeking streamlined automation, while Azure AI Language Studio NER caters to those with specific customization needs and expertise.

Regarding your question about algorithms - While the specific algorithms used were not be explicitly disclosed anywhere in the official documentation, both Custom NER and NER in AutomatedML likely leverage state-of-the-art machine learning techniques for NER tasks.

Please go through the below Azure documentation links, for relevant information:

Here is the document for custom NER using the Language Studio and REST API.

Here is the document for AutoML Text NER.

I hope you understand. Thank you.

---

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