Page 16 - MSDN Magazine, December 15, 2017

Data Sources
Ingest
Prepare
Analyze
Publish
Consume
Visual Studio Code Extension with Stack Overflow Bot
ACTION
Program Synthesis by Example (PROSE) enables the bot to gen- erate code by learning from examples provided by the user. This lets developers quickly use generated code to perform trans- formations of input and output strings. For example, the developer provides an input string like “Jane Doe” and specifies the output string as“Doe,J.”.PROSElearnsfromthis and generates the code that then performs the transformation.
In addition, data scientists can build custom deep learning models using Azure Machine Learning to automatically predict which Stack Overflow tags will be most rele- vant to the questions asked. This
Stack Overflow Content Stack Overflow Data Dump (Content Post)
DATA
Azure Blob Raw Storage
Azure Blob
Azure Functions
Azure Machine Learning
Cognitive Services
(Text, Analytics, OCR, Computer Vision, Bing Custom Search)
INTELLIGENCE
Bot Service
Docker Image
Automated Tag Prediction Model Contained in a Docker Image
Figure 8 Stack Overflow Bot Architecture
Services like OCR, Text Analytics, Computer Vision and the Lan- guage Understanding Intelligent Service (LUIS), as well as program synthesis capabilities from the PROSE SDK (microsoft.github.io/prose) that enables developers to add intelligence to bots. These ser- vices are used to power the Stack Overflow bot, which enables developers to get answers to questions within their development environment. You can see the Stack Overflow bot in action at bit.ly/2hqnEst, and access the sample code at bit.ly/2zIDbZ9.
The Stack Overflow bot leverages Azure Functions to power its dialog analysis. An intelligent Azure Function orchestrates the components used to analyze a user-uploaded screenshot, calling on OCR to extract text in the image and Text Analytics to identify key phrases. Figure 8 shows the Stack Overflow Bot architecture.
Resources
enables the Stack Overflow bot to expand the keywords used to perform keyword searches with Bing Search, helping developers find the answer they need faster.
Wrapping Up
In this article, you learned about the exciting capabilities offered by the Microsoft AI platform. From Cognitive Services that enable you to jumpstart using AI for building intelligent applications, to customizing state-of-the-state computer vision deep learning models, to building deep learning models of your own with Azure Machine Learning, the Microsoft AI platform equips developers with the tools they need. The AI platform is open and flexible, and empowers developers to choose the technology and deep learning framework best suited for their scenarios and skills.
This is the beginning of an intelligent revolution, and the Microsoft AI platform empowers you with exciting enterprise-ready services, infrastructure and tools to build intelligent, innovative applications. Get started today with microsoft.com/ai! n
Joseph sirosh is the corporate vice president of the Cloud AI Platform at Microsoft, where he leads the company’s enterprise AI strategy and products such as Azure Machine Learning, Azure Cognitive Services, Azure Search and the Bot Framework. He’s passionate about machine learning and its applications and has been active in the field since 1990. Sirosh holds a doctorate in computer science from the University of Texas at Austin and a BTech in computer science and engineering from the Indian Institute of Technology Chennai.
Wee hyong Tok is a principal data science manager of the Cloud AI Platform at Microsoft, where he leads AI Prototyping and Innovation. He has advised many Fortune 500 companies on data platform architectures, and using AI for their strategic initiatives. Wee Hyong holds a doctorate in computer science from the National University of Singapore.
Thanks to the following Microsoft technical expert for reviewing this article: Anand Raman, chief of staff of the Cloud AI Platform at Microsoft
You can learn more about what’s new with Azure Machine Learning at bit.ly/2zVdpUs. Also, there are lots of resources available to help you get started with Azure Machine Learning. Kick things off with the setup and installation guide at aka.ms/aml-blog-setup and a quick start sample from aka.ms/aml-blog-iris. As you explore Azure Machine Learning dive deeper into more in-depth tutorials. Check out a three-part Iris classification tutorial at bit.ly/2AwjSSN, as well as a detailed data wrangling tutorial at bit.ly/2yOrJxQ.
In addition, you can find sample code and detailed scenario walk-throughs on several very interesting use cases. These include:
• Aerial Image Classification: Distributed training and oper- ationalization of a land-usage classification model (bit.ly/2jf7y5o).
• Document Collection Analysis: Developing a robust model for text analytics (bit.ly/2ypziXr).
• Predictive Maintenance: Building an end-to-end predictive maintenance solution using PySpark (bit.ly/2zyBW0N).
• Energy Demand Time Series Forecasting: Forecasting energy demands to predict future loads on an electrical grid (bit.ly/2hps9mL).
The possibilities of using AI to solve exciting real-world problems are endless! I can’t wait to see how you use the Microsoft AI platform to create the next breakthrough intelligent solution.
12 msdn magazine
Artificial Intelligence