Azure Bot Services Guide

Quick start on how to create a bot in Azure.

Transcript

1. Quick Start Guide for Bots-as-a-Services using Microsoft Bot Framework Azure Bot Services Guide
2. Azure Bot Service Preview Develop your way Built in code editor </> Integrated chat window Quick start templates
3. Azure Bot Service Preview Direct Line support API Embedded web chat Cognitive Services Channel support
4. Azure Bot Service Preview Scale on demand Reduced dev ops Powered by Azure Functions Continuous deployment
5. Let’s Start Here’s what you need to build a bot: Azure account Web Browser Visual Studio or any code editor (Optional)
6. Go to portal.azure.com Login to your account
7. Default home page
8. Click New Search “bot service”
9. Choose “Bot Service”
10. Click “Create”
11. Fill in the info Tick “pin to dashboard” for easier navigation. Once done, click “Create”.
12. Deploying
13. Once deployed, it will automatically navigate to this page. Click “Create Microsoft App ID and password.
14. Click this to generate password
15. Copy the password
16. Click here to go back
17. Paste the password here
18. Choose the language Note: This guide will use C# as primary language
19. Choose the template Then create the bot
20. Bots Template Available Basic Form Proactive Language Understanding Question & Answer
21. Question & Answer Bots Create a Question & Answer bots with qnamaker.ai.
22. Sign in using the same account as Azure account
23. Click OK
24. Generating the bots, may take a few minutes
25. Predefined code Code that initialize the bots Chat with bots within development environment Log
26. Incoming messages are handled here Train our bots here
27. Go to qnamaker.ai Choose “My Services”
28. Click Continue Note: QnA Maker uses Microsoft Cognitive Services (Text Analytics) to understand the query, for now in preview it offers free up to 10,000 transactions a month.
29. Click this to edit
30. Click to add new pairs
31. Input the question and answer Once done, click “Save and retrain” You can test your knowledge base here
32. Testing Environment
33. The QnA pairs we made just now
34. You can input several alternatives for the question If answer is wrong, you can choose the appropriate one to train the bots
35. The alternative that just created Import your own database here
36. Database can be imported thru URL or file
37. Copy the URL
38. Paste here Once done, click “Save and retrain” Go back to Knowledge Base to check imported pairs
39. Successfully imported
41. Go back to “Test” to test the imported pairs Once done, click “Save and retrain” Finally, click “Publish”
42. Click “Publish”
43. Successfully deployed, and we can go back to Azure portal to edit the bots
44. Test the bots here
45. Activity type the bots received
46. You can change “Welcome” to anything you want. This is the reply when new users added into the conversation Once done, click “Save”
47. Go to dev.botframework.com to deploy bots Click “My bots”
48. Choose the bots that you just created
49. Edit the details of this bot Chat environment
50. Change name or upload an icon for your bot
51. Available channel to deploy your bots
52. Example for deploying the bot to skype
53. LUIS Bots Bots with natural language understanding, based on Cognitive Services – Language Understanding Intelligence Service (LUIS).
54. Introduction In this tutorial, we will build a bot that reply to users the image of traffic camera in Singapore. Data are obtained from mytransport.sg, provided by LTA. In order to proceed this tutorial, please obtain the Account Key from LTA by registering in mytransport.sg.
55. Click OK
56. Generating the bots, may take a few minutes
58. Code for handling conversation
59. Go to luis.ai, and login using the same account
60. Find the apps that has been created automatically, and click edit.
61. Click “+” to add new Intents Notes Intent is the intention that you want LUIS to learn, when the model is trained, LUIS will categorize the command to the Intents that you’ve set.
62. Intent Name Sample command under this intent Then click Save
63. Click Submit
64. Let’s add another one
65. Submitting the command Click “+” to add new Entities Note Entities is the object that LUIS will pick up from the Intents.
66. Name the Entity, then click save
67. Newly added Entities Click “Review labels” Note Now we want to label which item in the command belongs to “cameraNo”.
68. Submitted command Respective Intents
69. Click “1001” and select “cameraNo”
70. Then click “submit” Note Each entities will have one color to differentiate among other entities.
71. You can also choose pre-built entities
72. Once done, click “Train” After trained, click “Publish”
73. Click “Update published application”
74. We can test our model here
75. Top score Intent Extracted entities
76. LUIS Luis.ai is a machine learning tools built by Microsoft to understand the natural language. Users don’t have to deal with the machine learning model, just have to define the intents and entities for the particular bots. In next section, we will move into coding.
77. Add this to create the response for “Greeting” intent
78. Add this to create the response for “cameraImage” intent
79. Test here, we coded the response for “Greeting” intent to be “Hello World!” Click save after editing the code
80. JSON Now, we will get data from web, bind the data and output to users. The data that we obtained is in JSON format, and we have to bind the data before we can use it.
81. Add the library
82. Declare the JSON object that we are going to receive
83. Add a function to obtain the data Note You can obtain the Account Key from mytransport.sg
84. Add this to obtain the data when this intent is triggered. Note cameraID is the entities that we extract from the command.
85. Now we find the camera within the data Note CameraNo is ranging from 1001 – 9999, but not all camera number is available. We find the ranking of this camera here.
86. The case if we can’t find the camera.
87. The case we found the camera and reply with attachments
88. Complete code for the intent [LuisIntent("cameraImage")] public async Task cameraImage(IDialogContext context, LuisResult result) { //obtain the json data from LTA and extract entities Rootobject Traffic = await GetTrafficImages.GetImage(); string cameraID = ""; cameraID = result.Entities[0].Entity; //find the camera from data int cameraNo = 0; int maxCameraNo = Traffic.value.Count() - 1; while (cameraID != Traffic.value[cameraNo].CameraID) { if (cameraNo == maxCameraNo) break; else cameraNo += 1; } //construct response //case when can't find the camera if (cameraNo == maxCameraNo) { await context.PostAsync("I can't find this camera."); // context.Wait(MessageReceived); } //case when can find the camera else { string cameraImageURL = Traffic.value[cameraNo].ImageLink; var replyTraffic = context.MakeMessage(); replyTraffic.Attachments = new List<Attachment>() { new Attachment() { ContentUrl = cameraImageURL, ContentType = "image/png", Name = "Camera.png" } }; await context.PostAsync(replyTraffic); await context.PostAsync("Here's your image."); // context.Wait(MessageReceived); } }
89. Click “save” once done Try here after saved
90. Done That’s it! Congratulations on building your first language understanding bots! Refer to next slides on how to improve the language understanding model.
91. Click “Suggest” for what users have sent, and improve the model here
92. Choose the right intent and entities. Then train and publish your application.
93. Form Bots Bots that guide users to fill up a form through conversation.
94. Form Bots Form bots is simple and useful, in terms of performing a guided conversation with the users. Typically, this guided conversation is used when you need the users to answer a series of questions.
95. Choose “Form” and click create
97. Incoming messages are handled here
98. Handle the main dialog Initiate the form flow Create the form
99. Handle the conversation once form is done
100. Form is created here
101. Add 1 more question here
102. Add 1 choice question here List of possible options
103. Let’s remove CarOptions and ColorOptions Remove the questions for carOptions and ColorOptions
104. Add 1 questions after choice, to prompt user elaborate the details
105. Click “Save” Compilation succeeded
106. Let’s try in the “Chat”, initial the conversation with “Hello!”, then form will be prompted.
107. Handle conversation when done
108. Proactive Bots Bots that proactively engage users when given a context, and initiate the conversation.