This post is co-authored with Sara Kandil
Today, we are announcing a preview of new asynchronous (batch) APIs for Text Analytics and Text Analytics for health, which enable developers to apply Natural Language Processing (NLP) to even more scenarios so they can identify key phrases, entities and even personally identifiable information (PII).
Asynchronous Analyze API for Text Analytics
Text Analytics is a generally available Azure Cognitive Service that lets you discover insights in text using Natural Language Processing (NLP). The service helps you identify key phrases and entities (people, place, organization, event, date among others), recognize text that contains personal information (PII) and analyze sentiment (positive, neutral, or negative).
To date, customers have been using Text Analytics by making synchronous calls to the service’s REST API, client library SDK, or by using containers to run Text Analytics in their own environment. Today, we are introducing a new preview Analyze operation for users to analyze larger documents asynchronously combining multiple Text Analytics features in one call. This gives customers the flexibility to analyze more information, at once, when their applications don’t need a synchronous response. The new asynchronous Analyze operation for Text Analytics supports individual documents of up to 125k characters, and up to 25 documents in a request.
The Analyze operation preview supports key phrase extraction, named entity recognition and PII recognition and is available in 5 Azure regions (West US 2, East US2, West Europe, North Europe, and Central US). Support for the rest of the Text Analytics capabilities and additional regions is coming soon.
Asynchronous Analyze API for Text Analytics for health
We are also introducing a new asynchronous hosted API for Text Analytics for health. As a refresher, early this year (July), we announced a preview of Text Analytics for health, a capability for the healthcare industry, trained to extract insights from medical data. With Text Analytics for health, users can:
- Detect words and phrases mentioned in unstructured text as entities that are associated with semantic types in the healthcare and biomedical domain – such as diagnosis, medication name, symptom/sign, and more.
- Link entities to medical ontologies and domain-specific coding systems (for example, the Unified Medical Language System), and extract meaningful connections between concepts mentioned in text (for example, finding the relationship between a medication name and the dosage associated with it.)
- Detect negation of the different entities mentioned in the text.
Example of Text Analytics for health at work.
Previously, Text Analytics for health was only available for use via containers. This new API gives users the option to use the hosted service and avoid the heavy lifting of hosting containers unless they need to.
The hosted Text Analytics for health operation supports document sizes up to 5k characters and up to 10 documents in a single request. It is available for use in the West US 2, East US 2, Central US, North and West Europe regions.
In summary, Text Analytics is now more accessible with more ways to use the capabilities depending on your scenario. You can:
- Call the synchronous endpoints to use the Text Analytics features.
- Call the asynchronous Analyze API to process larger documents with multiple Text Analytics features in a single call.
- Call the hosted asynchronous Text Analytics for health API if your dataset that is being analyzed has clinical and biomedical documents.
- Use Text Analytics containers to host the endpoint in your own environments that meets your privacy and security requirements.
The new Text Analytics asynchronous APIs are available to use in Preview today. Please refer to our documentation to learn more and get started with these new APIs.