twitter-chatbot.md

### Lab - twitter chatbot setup

Now, a selection of the tweets can be directed to a chatbot service to trigger a response (based on the Watson Assitant service).

Add the following 4-node flow to the existing Twitter analysis flow:

Now, get hold of an existing Chatbot configuration; for this lab, we will be using one of the many Bot definitions at the IBM Bot Asset Exchange. Click this link to access the sample Bot.

Click the Get this bot option and save the resulting JSON string to a local file. This file will be imported into the Watson Assistant tool shortly.

This bot responds to queries/inputs with the lyrics and wisdom of renowned artist Jay-Z.

Yo will need to create an instance of the Watson Assistant service in the IBM Cloud catalog -

As before when creating services, leave the name to default, create the instance, and when presented with the service overview panels, use the Launch Tool option to begin creating a Conversation workspace. The workspace will be populated by importing the earlier JSON file.

Select the Create a Workspace option

And carefully NOT clicking on the Create option, select the Import workspace icon and navigate to your saved Jay-Z bot JSON file.

This should result in a new workspace being created, using the definitions in the JSON file

This will create the 3 main areas of chatbot server configuration:

1. Intents:
2. Entities:
3. Dialog:

Using the existing Node-RED flow, select messages with a particularly high emotional count in any of the tone category streams from the twitter analysis [Joy, Anger, Disgust, Sadness, Fear], to direct into the Chatbot, using the Watson Conversation node

You will need the credentials for the Watson Assistant instance, to plug into the node configuration menu. Either:

• make a connection between the Watson Assitant instance, and your Node-Red application, and after re-stage, the credentials will automatically populate the Watson Conversation nodes
• copy the credentials from the Watson Assitant instance, and apply directly to the node configuration, and use straight away (as you did previously for the Watson Tone Analyzer service)

Note - you also need the Watson Assistant workspace identifier to plug in to the node configuration

Select the Workspaces tab, to access the workspaces in this instance.

Click the details menu and select View details to view the Workspace ID - add this into the corresponding field in the Node-RED Conversation node configuration panel.

The output from the conversation is a JSON object in msg.payload - to extract the text and display in the debug pane, update the debug node to select just that part of the message:

Configure the change node to move the twitter content into the msg.payload property:

Last, but not least, configure the function node to select messages based on high emotional tone:

var tones = msg.response.document_tone.tone_categories[0].tones;
var pass_on = false;
tones.forEach(function(tone){
    if(tone["score"] > 0.75) {
        pass_on = true;
    }
});
if (pass_on) return msg;

And finally !!

Congratulations! - you have completed the process of generating responses to select tweets, using a basic chatbot configuration.