Cognitive Services August API Updates

8월 25, 2016에 게시됨

Senior Program Manager, Cognitive Services

Microsoft Cognitive Services is a collection of intelligence and knowledge APIs that enable developers to make their applications more intelligent, engaging and discoverable. Cognitive Services includes intelligent APIs that allow systems to see, hear, speak, understand and interpret needs using natural methods of communication; and knowledge APIs that bring the power of the web to developers. Below are the August updates for Microsoft Cognitive Services APIs.

 

   Computer Vision API

Now you can tag up to 2,000 recognized objects (animate and inanimate) within an image and caption in clear, written descriptions. The update includes improved celebrity recognition.

Improved tagging provides object recognition within images with a selection of over 2,000 objects, animals and people. Once identified, new image captioning capabilities provide clear written descriptions for the 2,000 recognizable images. Our new celebrity recognition model recognizes 200,000 celebrities from business, politics, sports, and entertainment around the world. Learn more about the Computer Vision API at Cognitive Services Computer Vision.

 

   Face API

The Face API now provides additional flexibility for matching and identifying faces.

 

New capabilities enable face to person authentication, finding similar faces, setting recognition confidence thresholds, and new list person groups.

 

Learn more about the Face API at Cognitive Services Face.

 

  Video API

You can now specify and limit locations in a video where you want to report motion, set sensitivity levels to control granularity, and detect the exact area where motion occurred.

 

  • Polygonal detection zones: Now you can narrow down the locations where motion is reported by specifying the areas of the video you care about.
  • Set sensitivity levels: Select between three levels—low, medium, and high—to control the granularity of motion you want detected.
  • Merge thresholds: Combine multiple motion events into one by specifying the timespan at which multiple reports of motion are merged into one event.
  • Motion location: When motion is detected, the API defines the exact area where the motion occurred by using a rectangular bounding box.

 

Learn more about the Video API at Cognitive Services Video.

 

  Speaker Recognition API

Speaker Recognition only needs 30 seconds of speech to identify speakers. 

 

Speaker Recognition previously required 60 seconds of speech for both enrollment and identifciation operations. Speaker Recognition API now accepts any length of speech, although it is recommended that users input 30 seconds for enrollment and 10 seconds for identification to get the best API experience.

 

Learn more about the Speaker Recognition API at Cognitive Services Speaker Recognition.

 

   Language Understanding Intelligence Service (LUIS) API

LUIS free transactions are decreasing

 

The number of free monthly LUIS transactions is changing from 100,000 to 10,000 transactions effective September 4, 2016.

 

Learn more about the Language Understanding Intelligence Service API at Cognitive Services LUIS.

  

   Text Analytics API

Text Analytics API now available in multiple languages.

 

Text Analytics API Sentiment analysis can now be performed in English, Spanish, French, and Portuguese; and key phrases can be extracted from text in English, Spanish, German, and Japanese. These features are in addition to the Text Analytics API’s ability to recognize the language of text in 120 languages—plus topic detection for English documents. More information on the Text Analytics API's language capabilities can be found at Text Analytics API Language Support

 

Learn more about the Text Analytics API at Cognitive Services Text Analytics.

 

 

   Academic Knowledge API

You can now make an unlimited number of transactions each month at the rate of 25 cents for every 1,000 transactions as part of a paid preview.

 

Academic Knowledge API users have only been able to use the free tier and make up to 10,000 transactions each month. Now with the paid tier, Academic Knowledge subscribers can make an unlimited number of transactions at the rate of $0.25 per 1000 transactions. 

 

Learn more about the Academic Knowledge API at Cognitive Services Academic Knowledge.

 

   Recommendations API

New capabilities have been added to the Recommendations API.

 

The Recommendations API, added the following capabilities:

  • New Recommendations UI allows you to train and test your models without having to write a single line of code.
  • New batch scoring allows you to get thousands of recommendations at once.
  • Support for business rules
  • Now you can enumerate and download usage and catalog files
  • Build metrics support to query the quality of recommendations models
  • Ranking build support to query the quality of item features in a recommendations model
  • Ability to search for a product in the catalog

 

Learn more about the Recommendations API at Cognitive Services Recommendations.