Introduction
Language is foundational to the development of human intelligence. Given language’s ubiquity, few areas of technology and artificial intelligence have had the opportunity to transform society at scale. In line with Microsoft’s mission to empower every person and every organization on the planet to achieve more, at Azure Cognitive Services, we continue to bring natural language processing services to customers around the world, breaking language barriers. Powered by state-of-the-art language models, our Language AI products empower you to extract, analyze, and retrieve meaningful insights from any form of text content.
Azure Cognitive Services Language products are created to offer flexibility for your business needs across many capabilities. Whether you are a high-growth start-up looking to improve team productivity by getting a concise summary of your customer calls right after the calls, a US-based company looking to understand the sentiment of your customers who speak Korean, or a contact center looking to enable intelligent, personalized conversational experiences.
We are excited to announce many new features and capabilities for Azure Cognitive Services for Language at this year’s Microsoft Ignite, including Conversation narrative summarization and chaptering, document abstractive summarization, enhanced contact center language AI capabilities, 90+ language support for Sentiment Analysis, Opinion Mining, and Key Phrase Extraction, 24+ language support for Named Entity Extraction, a new Document Translator Studio experience, support for asset versioning for custom text classification, custom NER and conversational language understanding, and dynamic model versioning to ease your model development process.
Drive business results faster through Text Summarization
With only 24 hours in a day, don’t you wish that everything at work had a TL; DR? Now imagine getting a TL; DR not just on a single document, but thousands! And what if you could get a concise summary for every conversation with chapters at the end of each call so that you don’t have to take copious notes and read everything to understand the key takeaways. Our new, expanded summarization capabilities for documents and conversations tackles this for you. Through unified language APIs for document abstractive summarization, conversation narrative summarization, and conversations chaptering, get quickly up to speed on all your work content and locate critical information in minutes. These three new preview features are powered by Z-Code++, our latest state-of-the-art pre-trained transformer language model.
Conversation narrative summarization and chaptering – These two new features can generate summaries and chapters for both chat logs and speech transcripts. Suitable for generic conversations and meetings, these features are an addition to the existing issue and resolution summarization for conversations product, suitable for customer support and call center use cases.
Document abstractive summarization – This new feature is an addition to the existing extractive summarization feature and provides natural language summaries of documents. These intelligent summaries collectively represent the most important or relevant information within the original content, instead of simply extracted sentences from the original document, enabling you to get the most valuable information from long text documents.
Learn more about these new summarization features and sign up for the gated preview access on this page.
Understand your customers wherever they are through expanded language support
Whether your company is a large multinational with a global presence, or a small-to-medium sized startup looking to win in international markets, understanding your customers, and delivering your solutions often involves processing and understanding customer inputs in many languages using many tools. We recognize that your users are becoming increasingly global, so we have expanded language support for multiple language skills
- Sentiment analysis will support 94 languages (starting Oct 15th)
- Key phrase extraction will support 94 languages (starting Oct 15th)
- Named entity recognition will support 24 languages (starting Oct 15th)
- Text Analytics for Health will support 7 languages (starting Oct 15th)
- Opinion Mining will support 94 languages (starting Nov 1st)
Boost your contact center operations through powerful AI capabilities
Great customer service experience is critical to a company’s success, impacting everything from sales volume to customer loyalty. To help you accelerate improvements to interactions between contact center agents and customers, we have enhanced our conversational product offerings, including Conversational language understanding that enables users to build custom natural language understanding models to predict the overall intention of an incoming utterance and Question answering to help you build conversational client applications, such as social media applications and chat bots. At Microsoft Build this year, we announced Conversation Personally Identifiable Information (PII) that enables you to extract sensitive information (PII) in conversations across several pre-defined categories and redact them, and issue and resolution summarization for conversations that produces a summary of issues and resolutions in transcripts of web chats and service call transcripts between customer-service agents, and your customers. To provide you with actionable customer insights from customer conversations, we continue to enhance the Sentiment Analysis capability to support conversation text. The new try-out experience for call center analytics in Language Studio will help you experience the power of our language models at the click of a few buttons.
Easily ingest and translate documents through the new document translator studio
Translation is at the heart of global communication, and documents contain a large portion of key information. Document translator makes it easy to translate most common document formats, including Word documents, PowerPoint presentations, PDF files, Excel sheets, and plain text. Translation is available for 100 languages through REST APIs and client-library SDKs. Starting November 1, you will be able to try out the new document translator experience on Language Studio. Simply connect your storage accounts, upload files, and specify translation parameters to achieve large-scale document translation without any coding or development.
New service capabilities
We continue to improve services for custom text classification, custom named entity recognition, and conversational language understanding (CLU) launched in June this year. We now provide asset versioning to keep track of the different data iterations that were used to train your models. We have also introduced multi-region deployment to deploy your models globally and access them in multiple regions. In addition, CLU has introduced regular expression entity components to extract custom patterns. The Named Entity Recognition service will also introduce entity resolution to return additional standard formats for quantified types, such as resolving “eighty seven” to the integer 87.
Updates in API calls with Generally Available (GA) model versions for pre-built capabilities
Today, static model versioning does not provide the flexibility to provide a dynamic AI solution. In addition, keeping track of model version numbers and model deprecation deadlines can be cumbersome. To address this, starting November 1, you will no longer need to specify a version number for GA models within an API call. Instead, you can select “latest” as the model-version parameter to ensure that you are always using our latest, highest quality models, without interrupting implementation. For preview models, you can continue to specify model version numbers. Different from model versions, API versions will still be supported with each API call to ensure reliability and stability of your applications. For more information about the updated model lifecycle policy and versioning experience, please see Language Model Lifecycle.
Want to learn more about driving better business results through Azure Cognitive Services for Language? Join us at Microsoft Ignite (Oct 12-14, 2022) and check out these additional resources
- Watch session topics featured in this blog
- Unlock new customer experiences with NLP at scale
- Explore Azure announcements at Microsoft Ignite