Today we are excited to introduce two new additions, summarization and PII detection/redaction for conversations. They are backed with the latest state-of-the-art transformer Z-Code models to empower users to analyze and unleash the potential in the wealth of conversational data. The two features can operate on both chat logs and speech transcripts, allowing seamless integration with Cognitive Service for Speech. An example and ready-for-use solution, Ingestion Client, enables end-to-end analysis of speech conversations, from audio to transcript to insights.
Conversation summarization – powered by the best-in-class Z-Codetrained with GPT-3, this first release focuses on the needs of customer support and call centers. Customer support agents typically spend 5 - 15 minutes writing notes when wrapping up each call or text chat, or when they transfer a case to the next level of support. This considerable time and effort significantly slow down the time to resolution. Our new feature automatically generates a summary of issues and resolutions from a two-party conversation, especially between a customer and an agent, which can greatly reduce case handling time, increase agents' job satisfaction, sustain high customer engagement, improve customer experiences, as a result boost customer loyalty. Built with this API offering in Azure Cognitive Services, Dynamics 365 Customer Service now enables this capability out-of-box for their customers
Conversation summarization is a gated public preview, and to request access please fill out this form.
Conversation PII – this feature detects and redacts sensitive Personally Identifiable Information (PII) in conversations. Companies have a wealth of conversational data that often contains PII. Before storing and using this data, PII masking is required due to Privacy Compliance laws and regulations. Detecting and redacting PII manually is a painful process that requires a huge amount of time and effort. The PII feature can help alleviate the encumbrance of processing PII. The first release of this feature recognizes entities including Credit Card, Name, and other frequently occurring types. The complete list of PII entities covered by the feature can be obtained here. The feature can also be used to redact audio segments in the original speech recording
The two features are now available in the following Azure regions:
In the first release, both features support English only. Z-Code model is born multi-lingual. We are expanding the language supporting list with validation.
The two new features are free while in preview. Pricing models will be defined when they are generally available.
APIs The Language APIs are the entry point to programmatically call all the Language features. Conversational features are available under /language/analyze-conversations. To access the two new features, please also specify the api-version, thus the URL will be like language/analyze-conversations/jobs?api-version=2022-05-15-preview