Announcing preview of 3 new summarization features in Cognitive Service for Language

Posted by

This post has been republished via RSS; it originally appeared at: Microsoft Tech Community - Latest Blogs - .

AT Microsoft Build 2022 in May, we announced the release of issue and resolution summarization for conversation. This expanded Microsoft support for summarization to cover both document and conversation as input genre, together with the in-production extractive summarization for document.


Today we are excited to introduce three new additions and key improvements for the two existing features. The three new features are document abstractive summarization, conversation narrative summarization, and conversations chaptering. They are also powered by Z-Code++ which is the latest state-of-the-art pre-trained transformer language model, to empower users to analyze and unleash the potential in the wealth of data, either documental or conversational.

The three new summarization features like the issue and resolution for conversations are in a gated public preview and to request access please fill out this form. The update for document extractive summarization will publicly be available in October.


Conversation narrative and chaptering The two new features can also operate on both chat logs and speech transcripts. However, different from the issue and resolution summarization for conversations, which focuses on the needs of customer support and call centers, these two new features are more suitable for generic conversations or meetings. For example, you can generate chapters for a recorded meeting for users to quickly navigate to the topic of interests without watching the whole meeting; you can quickly summarize discussion among participants in each chapter for users to easily get key talking points without notetaking.


Document abstractive summarization – This new feature generates a summary with concise, coherent sentences which are not extracted from the original document. It is an addition to the existing extractive summarization. So, now you can choose them based on your needs. For example, you can use both abstractive and extractive summarization to generate document summaries so that to enable readers to quickly find and prioritize documents without opening documents; you can use extractive summarization to extract salient sentences together their positioning information so that to allow readers to easily locate a summarized topic in a document and read it in the right context; you can use abstractive summarization to generate an ideal length of concise summary, which could be used for downstream information analytics or classification, et.



Regional availability

The three new summarization features are now available in the following Azure regions:

  • East US
  • North Europe
  • UK South

Supported Language

  • In the first release, the three new summarization features support English only. We are expanding the list of supported languages with validated quality.


The three new features are free during the initial preview. Pricing models will be defined and communicated when they are ready.


The Language APIs are the entry point to programmatically call all the Language features.

  • Conversational features are available under /language/analyze-conversations
  • Documental features are available under /language/analyze-text.

To access the three new features and the key improvements of the two existing features, please also specify the api-version, thus the URL will be like

  • language/analyze-conversations/jobs?api-version=2022-10-01-preview for conversational features
  • language/analyze-text/jobs?api-version=2022-10-01-preview for documental features

We provide QuickStart guidance (Conversation Summarization, Document Summarization ) to help you bootstrapped and get the journey started.

We are eager to see your businesses benefit from these summarization features and looking forward to your feedback to further drive improvements.
















Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.